title stringlengths 15 188 | abstract stringlengths 400 1.8k | introduction stringlengths 9 10.5k | content stringlengths 778 41.9k | abstract_len int64 400 1.8k | intro_len int64 9 10.5k | abs_len int64 400 1.8k |
|---|---|---|---|---|---|---|
Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection | An important task for designing QA systems is answer sentence selection (AS2): selecting the sentence containing (or constituting) the answer to a question from a set of retrieved relevant documents. In this paper, we propose three novel sentence-level transformer pre-training objectives that incorporate paragraph-leve... | Question Answering (QA) finds itself at the core of several commercial applications, for e.g., virtual assistants such as Google Home, Alexa and Siri. Answer Sentence Selection (AS2) is an important task for QA Systems operating on unstructured text such as web documents. When presented with a set of relevant documents... | Answer Sentence Selection (AS2) Earlier approaches for AS2 used CNNs Paragraph/Document-level Semantics Transformers for Long Inputs Longformer In this section we formally define the task of AS2. Given a question q and a set of answer candidates A={a 1 , . . ., a n }, the objective is to select the candidate ā ∈ A that... | 893 | 2,037 | 893 |
Style Transfer Through Back-Translation | Style transfer is the task of rephrasing the text to contain specific stylistic properties without changing the intent or affect within the context. This paper introduces a new method for automatic style transfer. We first learn a latent representation of the input sentence which is grounded in a language translation m... | Intelligent, situation-aware applications must produce naturalistic outputs, lexicalizing the same meaning differently, depending upon the environment. This is particularly relevant for language generation tasks such as machine translation This paper introduces a novel approach to transferring style of a sentence while... | Given two datasets 2 } which represent two different styles s 1 and s 2 , respectively, our task is to generate sentences of the desired style while preserving the meaning of the input sentence. Specifically, we generate samples of dataset X 1 such that they belong to style s 2 and samples of X 2 such that they belong ... | 815 | 1,317 | 815 |
Don't Say That! Making Inconsistent Dialogue Unlikely with Unlikelihood Training | Generative dialogue models currently suffer from a number of problems which standard maximum likelihood training does not address. They tend to produce generations that (i) rely too much on copying from the context, (ii) contain repetitions within utterances, (iii) overuse frequent words, and (iv) at a deeper level, co... | Open-ended tasks such as dialogue reveal a number of issues with current neural text generation methods. In more strongly grounded tasks such as machine translation and image captioning, current encoder-decoder architectures provide strong performance, where mostly word-level decisions are often taken correctly by the ... | Dialogue Generation Dialogue generation consists in predicting an utterance y = (y 1 , . . . , y |y| ) given a context x = {s 1 , . . . , s k , u 1 , . . . , u t } that consists of initial context sentences s 1:k (e.g., scenario, knowledge, personas, etc.) followed by dialogue history utterances u 1:t from speakers who... | 964 | 1,409 | 964 |
Automatic Metric Validation for Grammatical Error Correction | Correction (GEC) is currently done by observing the correlation between human and metric-induced rankings. However, such correlation studies are costly, methodologically troublesome, and suffer from low inter-rater agreement. We propose MAEGE, an automatic methodology for GEC metric validation, that overcomes many of t... | Much recent effort has been devoted to automatic evaluation, both within GEC Human rankings are often considered as ground truth in text-to-text generation, but using them reliably can be challenging. Other than the costs of compiling a sizable validation set, human rank-ings are known to yield poor inter-rater agreeme... | We turn to presenting the metrics we experiment with. The standard practice in GEC evaluation is to define differences between the source and a correction (or a reference) as a set of edits BLEU. BLEU GLEU. GLEU iBLEU. iBLEU We set α = 0.8 as suggested by Sun and Zhou. F -Score computes the overlap of edits to the sour... | 672 | 2,406 | 672 |
Robust Hate Speech Detection via Mitigating Spurious Correlations | We develop a novel robust hate speech detection model that can defend against both wordand character-level adversarial attacks. We identify the essential factor that vanilla detection models are vulnerable to adversarial attacks is the spurious correlation between certain target words in the text and the prediction lab... | Online social media bring people together and encourage people to share their thoughts freely. However, it also allows some users to misuse the platforms to promote the hateful language. As a result, hate speech, which "expresses hate or encourages violence towards a person or group based on characteristics such as rac... | A hate speech detection model can be defined as a functional mapping from T to Y , where t ∈ T is a set of input texts and y ∈ Y is the target label set. In general, the output of the detection model is the softmax probability of predicting each class k, i.e., f k (t; θ) = P (Y = y k |t), where θ is the parameters of t... | 717 | 913 | 717 |
Named Entity Recognition with Character-Level Models | We discuss two named-entity recognition models which use characters and character ¤ -grams either exclusively or as an important part of their data representation. The first model is a character-level HMM with minimal context information, and the second model is a maximum-entropy conditional markov model with substanti... | For most sequence-modeling tasks with word-level evaluation, including named-entity recognition and part-ofspeech tagging, it has seemed natural to use entire words as the basic input features. For example, the classic HMM view of these two tasks is one in which the observations are words and the hidden states encode c... | Figure When using character-level models for word-evaluated tasks, one would not want multiple characters inside a single word to receive different labels. This can be avoided in two ways: by explicitly locking state transitions inside words, or by careful choice of transition topology. In our current implementation, w... | 565 | 888 | 565 |
A Formal Hierarchy of RNN Architectures | We develop a formal hierarchy of the expressive capacity of RNN architectures. The hierarchy is based on two formal properties: space complexity, which measures the RNN's memory, and rational recurrence, defined as whether the recurrent update can be described by a weighted finite-state machine. We place several RNN va... | While neural networks are central to the performance of today's strongest NLP systems, theoretical understanding of the formal properties of different kinds of networks is still limited. It is established, for example, that the Elman (1990) RNN is Turing-complete, given infinite precision and computation time Recently,... | We introduce a unified hierarchy (Figure We provide the first formal proof that LSTMs can encode functions that rational recurrences cannot. On the other hand, we show that the saturated Elman RNN and GRU are rational recurrences with constant space complexity, whereas the QRNN has unbounded space complexity. We also s... | 667 | 683 | 667 |
A Simple and Effective Approach to Coverage-Aware Neural Machine Translation | We offer a simple and effective method to seek a better balance between model confidence and length preference for Neural Machine Translation (NMT). Unlike the popular length normalization and coverage models, our model does not require training nor reranking the limited n-best outputs. Moreover, it is robust to large ... | In the past few years, Neural Machine Translation (NMT) has achieved state-of-the-art performance in many translation tasks. It models the translation problem using neural networks with no assumption of the hidden structures between two languages, and learns the model parameters from bilingual texts in an end-to-end fa... | Given a word sequence, a coverage vector indicates whether the word of each position is translated. This is trivial for statistical machine translation However, it is not the case for NMT where the coverage is modeled in a soft way. In NMT, no explicit translation units or rules are used. The attention mechanism is use... | 522 | 2,040 | 522 |
Deep-speare: A joint neural model of poetic language, meter and rhyme | In this paper, we propose a joint architecture that captures language, rhyme and meter for sonnet modelling. We assess the quality of generated poems using crowd and expert judgements. The stress and rhyme models perform very well, as generated poems are largely indistinguishable from human-written poems. Expert evalua... | With the recent surge of interest in deep learning, one question that is being asked across a number of fronts is: can deep learning techniques be harnessed for creative purposes? Creative applications where such research exists include the composition of music (Humphrey et al., 2013; Sturm et al., 2016; Choi et al., ... | Early poetry generation systems were generally rule-based, and based on rhyming/TTS dictionaries and syllable counting (Gervás, 2000; Wu et al., 2009; Netzer et al., 2009; Colton et al., 2012; Toivanen et al., 2013). The earliest attempt at using statistical modelling for poetry generation was Greene et al. (2010), b... | 649 | 1,971 | 649 |
Evolutionary Data Measures: Understanding the Difficulty of Text Classification Tasks | Classification tasks are usually analysed and improved through new model architectures or hyperparameter optimisation but the underlying properties of datasets are discovered on an ad-hoc basis as errors occur. However, understanding the properties of the data is crucial in perfecting models. In this paper we analyse e... | If a machine learning (ML) model is trained on a dataset then the same machine learning model on the same dataset but with more granular labels will frequently have lower performance scores than the original model (see results in Such a difficulty measure would be useful as an analysis tool and as a performance estimat... | One source of difficulty in a dataset is mislabelled items of data (noise). Class Diversity. Class diversity provides information about the composition of a dataset by measuring the relative abundances of different classes Class Balance. Unbalanced classes are a known problem in machine learning Data Complexity. Humans... | 1,080 | 715 | 1,080 |
Neural Readability Pairwise Ranking for Sentences in Italian Administrative Language | Automatic Readability Assessment aims at assigning a complexity level to a given text, which could help improve the accessibility to information in specific domains, such as the administrative one. In this paper, we investigate the behavior of a Neural Pairwise Ranking Model (NPRM) for sentence-level readability assess... | Due to its complexity, the style of Italian administrative texts has been defined as "artificial" and "obscure" One way to tackle this problem is with technologies for Automatic Readability Assessment (ARA) that predict the complexity of texts In this paper, we tackle the data scarcity issue in two ways. First, we intr... | Early ARA techniques consisted in the so-called "readability formulae". Such formulae were created for educational purposes and mainly considered shallow text features, like word and sentence length or lists of common words However, longer words and sentences are not necessarily complex, and these formulae have been pr... | 1,114 | 1,897 | 1,114 |
A Human-Centric Evaluation Platform for Explainable Knowledge Graph Completion | Explanations for AI are expected to help human users understand AI-driven predictions. Evaluating plausibility, the helpfulness of the explanations, is therefore essential for developing eXplainable AI (XAI) that can really aid human users. Here we propose a human-centric evaluation platform 1 to measure plausibility o... | A Knowledge Graph (KG) is a structured representation of knowledge that captures the relationships between entities. It is composed of triples in the format (subject, relation, object), denoted as t = (s, r, o), where two entities are connected by a specified relation. For example, in the triple (London, isCapitalOf, U... | We build an online system to evaluate XKGCs in a human centric manner. Our system considers the real needs and interests of human users in collaboration with AI, allowing us to investigate: can humans assess correctness of a KGC prediction based on its explanations? Which explanations are helpful for human users? The a... | 721 | 2,363 | 721 |
Privacy Implications of Retrieval-Based Language Models | Retrieval-based language models (LMs) have demonstrated improved interpretability, factuality, and adaptability compared to their parametric counterparts by incorporating retrieved text from external datastores. While it is well known that parametric models are prone to leaking private data, it remains unclear how the ... | Retrieval-based language models | Email: mailme@alice.com mailme@bob.com mailme@charlie.com … URL: alice@bob.com harry@hogwarts.edu … text passages that are most relevant to the prompt provided to the model. These retrieved results are then utilized as additional information when generating the model's response to the prompt. Retrievalbased language mo... | 1,389 | 31 | 1,389 |
MULTIHIERTT: Numerical Reasoning over Multi Hierarchical Tabular and Textual Data | Numerical reasoning over hybrid data containing both textual and tabular content (e.g., financial reports) has recently attracted much attention in the NLP community. However, existing question answering (QA) benchmarks over hybrid data only include a single flat table in each document and thus lack examples of multist... | In recent years, as key to many NLP tasks such as QA, there is a flurry of works on numerical reasoning over various types of data including textual data However, existing QA datasets over hybrid data only contain a single flat table in each document To address these shortcomings, we present MUL-TIHIERTT: an expert-ann... | Question Answering Benchmark There are numerous QA datasets focusing on text, table/knowledge base (KB), and hybrid data. SQuAD Numerical Reasoning Numerical reasoning plays an important role in different NLP tasks Financial NLP Financial NLP has attracted much attention recently. There have been various application in... | 1,501 | 2,165 | 1,501 |
Sentiment Analysis on Streaming User Reviews via Dual-Channel Dynamic Graph Neural Network | Sentiment analysis on user reviews has achieved great success thanks to the rapid growth of deep learning techniques. The large number of online streaming reviews also provides the opportunity to model temporal dynamics for users and products on the timeline. However, existing methods model users and products in the re... | Sentiment analysis on user reviews, inferring the overall sentiment polarity (e.g. 1-5 stars on the review site Amazon) of a user-written review document for a product, has gained popularity with the rapid growth of online review sites such as Amazon, Yelp, and IMDB. Compared to other sentiment analysis tasks For senti... | Incorporating user and product information into models through reviews is the main idea of user review sentiment analysis methods. Recently, to exploit more knowledge from useruser or user-product relations, Amplayo et al. (2018) introduces shared vectors that are constructed from similar users/products to address the ... | 1,097 | 1,918 | 1,097 |
Combining the Best of Two Worlds: A Hybrid Approach to Multilingual Coreference Resolution | We describe our system for the CoNLL-2012 shared task, which seeks to model coreference in OntoNotes for English, Chinese, and Arabic. We adopt a hybrid approach to coreference resolution, which combines the strengths of rule-based methods and learningbased methods. Our official combined score over all three languages ... | The CoNLL-2012 shared task extends last year's task on coreference resolution from a monolingual to a multilingual setting Our decision to adopt a hybrid approach is motivated by the observation that rule-based methods and learning-based methods each have their unique strengths. As shown by the Stanford coreference res... | To build a mention detector that strikes a relatively good balance between precision and recall, we employ a two-step approach. First, in the extraction step, we identify named entities (NEs) and employ language-specific heuristics to extract mentions from syntactic parse trees, aiming to increase our upper bound on re... | 421 | 863 | 421 |
Question-Answering in a Low-resourced Language: Benchmark Dataset and Models for Tigrinya | Question-Answering (QA) has seen significant advances recently, achieving near human-level performance over some benchmarks. However, these advances focus on high-resourced languages such as English, while the task remains unexplored for most other languages, mainly due to the lack of annotated datasets. This work pres... | Question Answering (QA) and Machine Reading Comprehension (MRC) have seen significant advances in recent years, achieving human-level performance on large-scale benchmarks This work presents TiQuAD, the first publicly available Question-Answering Dataset for Tigrinya; see Figure We assess the quality of annotations and... | Native reading comprehension datasets beyond the English language are relatively rare. Efforts have been made to build MRC datasets in Chinese, French, German, and Korean, among others, all of which are designed following the formulation of SQuAD. The SberQuAD dataset Cross-lingual Question Answering Languagespecific d... | 1,096 | 1,373 | 1,096 |
Coreference in Wikipedia: Main Concept Resolution | Wikipedia is a resource of choice exploited in many NLP applications, yet we are not aware of recent attempts to adapt coreference resolution to this resource. In this work, we revisit a seldom studied task which consists in identifying in a Wikipedia article all the mentions of the main concept being described. We sho... | Coreference Resolution (CR) is the task of identifying all mentions of entities in a document and grouping them into equivalence classes. CR is a prerequisite for many NLP tasks. For example, in Open Information Extraction (OIE) Most CR systems, including state-of-the-art ones It is now widely accepted that coreference... | Our approach is inspired by, and extends, previous works on coreference resolution which show that incorporating external knowledge into a CR system is beneficial. In particular, a variety of approaches One issue with all the aforementioned studies is that inaccuracies often cause cascading errors in the pipeline Deali... | 792 | 2,017 | 792 |
Task-oriented Dialogue System for Automatic Diagnosis | In this paper, we make a move to build a dialogue system for automatic diagnosis. We first build a dataset collected from an online medical forum by extracting symptoms from both patients' self-reports and conversational data between patients and doctors. Then we propose a taskoriented dialogue system framework to make... | Automatic phenotype identification using electronic health records (EHRs) has been a rising topic in recent years In general, each EHR contains multiple types of data, including personal information, admission note, diagnose tests, vital signs and medical image. And it is collected accumulatively following a diagnostic... | Our dataset is collected from the pediatric department in a Chinese online healthcare community For each patient, we can also obtain the final diagnosis from doctors as the label. For clarity, we term symptoms from self-reports as explicit symptoms while those from conversational data as implicit symptoms. We choose fo... | 708 | 1,681 | 708 |
Learning to Ignore Adversarial Attacks | Despite the strong performance of current NLP models, they can be brittle against adversarial attacks. To enable effective learning against adversarial inputs, we introduce the use of rationale models that can explicitly learn to ignore attack tokens. We find that the rationale models can successfully ignore over 90% o... | Adversarial robustness is an important issue in NLP, asking how to proof models against confounding tokens designed to maliciously manipulate model outputs. As such models become more powerful and ubiquitous, research continues to discover surprising vulnerabilities A common defense method to combat adversarial attacks... | We build on prior work on adversarial robustness and learning from explanations. Adversarial robustness. Adversarial attacks against NLP models seek to maliciously manipulate model output by perturbing model input. As interest in adversarial attacks has increased, so has interest in developing models robust to these at... | 750 | 2,088 | 750 |
Directions for NLP Practices Applied to Online Hate Speech Detection | Addressing hate speech in online spaces has been conceptualized as a classification task that uses Natural Language Processing (NLP) techniques. Through this conceptualization, the hate speech detection task has relied on common conventions and practices from NLP. For instance, inter-annotator agreement is conceptualiz... | Online hate speech is the cause for growing concern, due to its social impacts The increase in research interest to hate speech detection has spurred on a growth and variety in annotated resources for the task created within the academy and industry. However, at the same time, critical work on hate speech detection has... | Hate speech detection is commonly conceptualized as a supervised classification task, with the goal to determine whether content is hateful or not Defining hate speech is to control the discourse surrounding the phenomena; determine which groups are minoritized, and therefore should be protected; and which patterns of ... | 800 | 1,393 | 800 |
A Simple and Effective Usage of Word Clusters for CBOW Model | We propose a simple and effective method for incorporating word clusters into the Continuous Bag-of-Words (CBOW) model. Specifically, we propose to replace infrequent input and output words in CBOW model with their clusters. The resulting cluster-incorporated CBOW model produces embeddings of frequent words and a small... | Word embeddings have been widely applied to various natural language processing (NLP) tasks. These embeddings can be pretrained on a large corpus and carry useful semantic information. One of the most well-known methods for obtaining word embeddings is based on Continuous Bag-of-Words (CBOW) In this paper, we focus on ... | A number of related research efforts have been done to help to learn better word embeddings aiming at different aspects. For example, There have also been some previous researches that utilized word clusters for reducing the number of word embeddings. 3 Our Method Let w t denote the t-th word in a given text. We adopt ... | 600 | 1,425 | 600 |
Aligning Medical Domain Ontologies for Clinical Query Extraction | Often, there is a need to use the knowledge from multiple ontologies. This is particularly the case within the context of medical imaging, where a single ontology is not enough to provide the complementary knowledge about anatomy, radiology and diseases that is required by the related applications. Consequently, semant... | As the content of numerous ontologies in the biomedical domain increases, so does the need for sharing and reusing this body of knowledge. Often, there is a need to use the knowledge from multiple ontologies. This is particularly the case within the context of medical imaging, where a single ontology is not enough to s... | The following scenario illustrates how the alignment of medical ontologies facilitates the integration of medical knowledge that is relevant to medical image contents from multiple ontologies. Suppose that we want to help a radiologist, who searches for related information about the manifestations of a certain type of ... | 916 | 3,019 | 916 |
Training Adaptive Computation for Open-Domain Question Answering with Computational Constraints | Adaptive Computation (AC) has been shown to be effective in improving the efficiency of Open-Domain Question Answering (ODQA) systems. However, current AC approaches require tuning of all model parameters, and training state-of-the-art ODQA models requires significant computational resources that may not be available f... | Open-Domain Question Answering (ODQA) requires finding relevant information for a given question and aggregating the information to produce an answer. The retriever-reader architecture, popularised by In this work, we explore an efficient approach to apply adaptive computation to large generative ODQA models. We introd... | Open Domain Question Answering ODQA is a task that aims to answer a factoid question given a document corpus. Most works in this domain follow a retriever-reader design first proposed by However, thanks to recent advances in sequenceto-sequence pretrained language models Adaptive Computation Adaptive computation allows... | 965 | 853 | 965 |
Learning to Translate in Real-time with Neural Machine Translation | Translating in real-time, a.k.a. simultaneous translation, outputs translation words before the input sentence ends, which is a challenging problem for conventional machine translation methods. We propose a neural machine translation (NMT) framework for simultaneous translation in which an agent learns to make decision... | Simultaneous translation, the task of translating content in real-time as it is produced, is an important tool for real-time understanding of spoken lectures or conversations In this paper, we propose a unified design for learning to perform neural simultaneous machine translation. The proposed framework is based on fo... | Suppose we have a buffer of input words X = {x 1 , ..., x Ts } to be translated in real-time. We define the simultaneous translation task as sequentially making two interleaved decisions: READ or WRITE. More precisely, the translator READs a source word x η from the input buffer in chronological order as translation co... | 720 | 1,096 | 720 |
Punctuation: Making a Point in Unsupervised Dependency Parsing | We show how punctuation can be used to improve unsupervised dependency parsing. Our linguistic analysis confirms the strong connection between English punctuation and phrase boundaries in the Penn Treebank. However, approaches that naively include punctuation marks in the grammar (as if they were words) do not perform ... | Unsupervised dependency parsing is a type of grammar induction -a central problem in computational linguistics. It aims to uncover hidden relations between head words and their dependents in free-form text. Despite decades of significant research efforts, the task still poses a challenge, as sentence structure is under... | Frac Punctuation and syntax are related But are there simple enough connections between the two to aid in grammar induction? This section explores the regularities. Our study of punctuation in WSJ Out of 51,558 sentences, most -37,076 (71.9%)contain sentence-internal punctuation. These punctuated sentences contain 123,... | 1,097 | 1,387 | 1,097 |
Polyglot Prompting: Multilingual Multitask Prompt Training | This paper aims for a potential architectural improvement for multilingual learning and asks: Can different tasks from different languages be modeled in a monolithic framework, i.e. without any task/language-specific module? The benefit of achieving this could open new doors for future multilingual research, including ... | The emergence of multilingual pre-trained language models sated through the higher-resource languages shared with them. Despite the preliminary success in the lowresource scenarios using shared knowledge across languages in multilingual language models Unifying different tasks into one framework can be challenging if w... | Multitask & Multilingual Learning The developments of neural networks have made it easier to share information across tasks or languages. As such, in the past few years, there has been much work on multitask learning within the same language Prompting Methods Prompting is a technique that aims to make better use of pre... | 1,237 | 1,395 | 1,237 |
QASR: QCRI Aljazeera Speech Resource A Large Scale Annotated Arabic Speech Corpus | We introduce the largest transcribed Arabic speech corpus, QASR 1 , collected from the broadcast domain. This multi-dialect speech dataset contains 2, 000 hours of speech sampled at 16kHz crawled from Aljazeera news channel. The dataset is released with lightly supervised transcriptions, aligned with the audio segments... | Research on Automatic Speech Recognition (ASR) has attracted a lot of attention in recent years Natural Language Processing (NLP), on the other hand values large amount of textual information for designing experiments. NLP research for Arabic has achieved a milestone in the last few years in morphological disambiguatio... | The CallHome task within the NIST benchmark evaluations framework The following datasets are released from the Multi-Genre Broadcast MGB challenge: (i) MGB-2 We obtained Aljazeera Arabic news channel's archive (henceforth AJ), spanning over 11 years from 2004 until 2015. It contains more than 4, 000 episodes from 19 d... | 1,238 | 1,906 | 1,238 |
Learning to Recover from Multi-Modality Errors for Non-Autoregressive Neural Machine Translation | Non-autoregressive neural machine translation (NAT) predicts the entire target sequence simultaneously and significantly accelerates inference process. However, NAT discards the dependency information in a sentence, and thus inevitably suffers from the multi-modality problem: the target tokens may be provided by differ... | Although neural machine translation (NMT) has achieved state-of-the-art performance in recent years Recently, non-autoregressive neural machine translation (NAT) models Intensive efforts have been devoted to alleviate the above problem, which can be roughly divided into two lines. The first line of work leverages the i... | Autoregressive neural machine translation (AT) generates the translation token-by-token conditioned on translation history. Denoting a source sentence as x = {x i } T i=1 and a target sentence as y = {y j } T j=1 , AT models the joint probability as: where y <t denotes the generated tokens before y t . During decoding,... | 1,323 | 2,616 | 1,323 |
Factored Statistical Machine Translation for Grammatical Error Correction | This paper describes our ongoing work on grammatical error correction (GEC). Focusing on all possible error types in a real-life environment, we propose a factored statistical machine translation (SMT) model for this task. We consider error correction as a series of language translation problems guided by various lingu... | The task of grammatical error detection and correction (GEC) is to make use of computational methods to fix the mistakes in a written text. It is useful in two aspects. For a non-native English learner it may help to improve the grammatical quality of the written text. For a native speaker the tool may help to remedy m... | In contrast with phrase-based translation models, factored models make use of additional linguistic clues to guide the system such that it generates translated sentences in which morphological and syntactic constraints are met In order to construct a SMT model, we convert the training data into a parallel corpus where ... | 983 | 1,785 | 983 |
PAIRSPANBERT: An Enhanced Language Model for Bridging Resolution | We present PAIRSPANBERT, a SPANBERTbased pre-trained model specialized for bridging resolution. PAIRSPANBERT is pre-trained with a novel objective that aims to learn the contexts in which two mentions are implicitly linked to each other from a large amount of data automatically generated either heuristically or via dis... | Bridging is essential for establishing coherence among the entities within a text through nonidentical semantic or encyclopedic relations (1) In June, farmers held onto meat, milk and grain, waiting for July's usual state directed price rises. The Communists froze prices instead. The task of bridging resolution, which ... | Bridging resolution. The two sub-tasks of bridging resolution, namely bridging anaphora recognition and bridging anaphora resolution, have been tackled separately. One line of research has modeled bridging anaphora recognition as a part of the information status (IS) classification problem where each discourse entity i... | 672 | 2,404 | 672 |
Style Transfer as Data Augmentation: A Case Study on Named Entity Recognition | In this work, we take the named entity recognition task in the English language as a case study and explore style transfer as a data augmentation method to increase the size and diversity of training data in low-resource scenarios. We propose a new method to effectively transform the text from a high-resource domain to... | Large-scale pre-trained language models (PLMs) such as BERT Data augmentation is effective in addressing data scarcity. Previous work In this work, we explore the potential of employing style transfer as a way of data augmentation in cross-domain settings. Style transfer on natural language aims to change the style-rel... | Style Transfer Style transfer aims to adjust the stylistic characteristics of a sentence while preserving its original meaning. It has been widely studied in both supervised To facilitate research in this direction, we study style transfer as data augmentation and propose a novel approach to explore transferring the st... | 912 | 1,265 | 912 |
A Bayesian mixture model for term re-occurrence and burstiness | This paper proposes a model for term reoccurrence in a text collection based on the gaps between successive occurrences of a term. These gaps are modeled using a mixture of exponential distributions. Parameter estimation is based on a Bayesian framework that allows us to fit a flexible model. The model provides measure... | Traditionally, Information Retrieval (IR) and Statistical Natural Language Processing (NLP) applications have been based on the "bag of words" model. This model assumes term independence and homogeneity of the text and document under consideration, i.e. the terms in a document are all assumed to be distributed homogene... | Previous attempts to model a term's distribution pattern have been based on the Poisson distribution. If the number of occurrences of a term in a document is denoted by k, then the model assumes: for k = 0, 1, 2, . . . Estimates based on this model are good for non-content, non-informative terms, but not for the more i... | 605 | 2,623 | 605 |
Your Answer is Incorrect... Would you like to know why? Introducing a Bilingual Short Answer Feedback Dataset | Handing in a paper or exercise and merely receiving "bad" or "incorrect" as feedback is not very helpful when the goal is to improve. Unfortunately, this is currently the kind of feedback given by many Automatic Short Answer Grading (ASAG) systems. One of the reasons for this is a lack of content-focused elaborated fee... | Assessment and feedback are essential to highquality education Besides being cost-and time-efficient, automating assessment also offers unique teaching opportunities. As long as systems give individual, responsespecific feedback, learners may retry or take additional assignments and receive instantaneous feedback as of... | What are extension headers in IPv6 and where are they located in a packet? What is the main advantage of extension headers compared to IPv4? Reference Answer: Extension headers are used to extend the fixed IPv6 header with additional, optional network layer information. If present, they are located between the fixed he... | 1,070 | 696 | 1,070 |
Extracting Commonsense Properties from Embeddings with Limited Human Guidance | Intelligent systems require common sense, but automatically extracting this knowledge from text can be difficult. We propose and assess methods for extracting one type of commonsense knowledge, object-property comparisons, from pretrained embeddings. In experiments, we show that our approach exceeds the accuracy of pre... | Automatically extracting common sense from text is a long-standing challenge in natural language processing In this paper, we study methods for reducing the amount of human input needed to learn common sense. Specifically, we focus on learning relative comparisons of (one-dimensional) object properties, such as the fac... | We define the task of comparing object properties in two different ways: a three-way classification task, and a four-way classification task. In the three-way classification task, we want to estimate the following conditional probability: For example, P rob(An elephant is larger than a dog) can be expressed as P (L = >... | 496 | 534 | 496 |
Kandinsky: an Improved Text-to-Image Synthesis with Image Prior and Latent Diffusion | Text-to-image generation is a significant domain in modern computer vision and has achieved substantial improvements through the evolution of generative architectures. Among these, there are diffusion-based models that have demonstrated essential quality enhancements. These models are generally split into two categorie... | In quite a short period of time, generative abilities of text-to-image models have improved substantially, providing users with photorealistic quality, near real-time inference speed, a great number of applications and features, including simple easyto-use web-based platforms and sophisticated AI graphics editors. This... | Early text-to-image generative models, such as DALL-E This enabled a wide array of applications like 3D object synthesis Diffusion models achieve state-of-the-art results in image generation task both unconditional Text-to-image diffusion models have become a popular research direction due to the high performance of di... | 1,300 | 1,901 | 1,300 |
Adaptive Gating in Mixture-of-Experts based Language Models | Large language models, such as OpenAI's Chat-GPT, have demonstrated exceptional language understanding capabilities in various NLP tasks. Sparsely activated mixture-of-experts (MoE) has emerged as a promising solution for scaling models while maintaining a constant number of computational operations. Existing MoE model... | The field of natural language processing (NLP) has undergone a remarkable revolution driven by the rapid advancements in language models (Cha; Sparsely-activated mixture-of-experts (MoE) is a promising paradigm to address the scalability issue while maintaining a constant number of computation FLOPs Towards this end, w... | 2.1 Mixture-of-Experts Mixture-of-Experts (MoE) has been adopted in various deep neural network models In particular, these models typically employ an MoE layer to substitute the feed-forward network (FFN) layer. The MoE layer comprises multiple FFNs, each acting as an expert, along with a gating network. Each expert i... | 1,313 | 3,308 | 1,313 |
Restricted Recurrent Neural Tensor Networks: Exploiting Word Frequency and Compositionality | Increasing the capacity of recurrent neural networks (RNN) usually involves augmenting the size of the hidden layer, with significant increase of computational cost. Recurrent neural tensor networks (RNTN) increase capacity using distinct hidden layer weights for each word, but with greater costs in memory usage. In th... | Recurrent neural networks (RNN), which compute their next output conditioned on a previously stored hidden state, are a natural solution to sequence modeling. In this paper, we propose the Restricted RNTN (r-RNTN) which uses only K < |V | recurrence matrices. Given that |V | words must be assigned K matrices, we map th... | We focus on related work that addresses language modeling via RNNs, word representation, and conditional computation. Given a sequence of words (x 1 , ..., x T ), a language model gives the probability P (x t |x 1...t-1 ) for t ∈ [1, T ]. Using a RNN, where where i(z) maps a hot-one encoded vector to its integer repres... | 801 | 698 | 801 |
Knowledge Graph-Augmented Abstractive Summarization with Semantic-Driven Cloze Reward | Sequence-to-sequence models for abstractive summarization have been studied extensively, yet the generated summaries commonly suffer from fabricated content, and are often found to be near-extractive. We argue that, to address these issues, the summarizer should acquire semantic interpretation over input, e.g., via str... | Abstractive summarization aims to produce concise and informative summaries with the goal of promoting efficient information consumption and knowledge acquisition | The Week column. Mayor John Fabrizi of Brigeport, Conn, publicly admits he used cocaine and abused alcohol while in office; says he stopped drinking alcohol and sought help for his drug problem about 18 months ago. learning objectives, these models frequently produce unfaithful content To this end, we present ASGARD, a... | 1,248 | 162 | 1,248 |
Unsupervised Dual Paraphrasing for Two-stage Semantic Parsing | One daunting problem for semantic parsing is the scarcity of annotation. Aiming to reduce nontrivial human labor, we propose a two-stage semantic parsing framework, where the first stage utilizes an unsupervised paraphrase model to convert an unlabeled natural language utterance into the canonical utterance. The downst... | Semantic parsing is the task of converting natural language utterances into structured meaning representations, typically logical forms b). Researchers use crowdsourcing to paraphrase those canonical utterances into natural language utterances (the upper part of Figure Annotators may struggle to understand the exact me... | For the rest of our discussion, we use x to denote natural language utterance, z for canonical utterance, and y for logical form. X , Z and Y represent the set of all possible natural language utterances, canonical utterances, and logical forms respectively. The underlying mapping function f : Z -→ Y is dominated by gr... | 779 | 2,170 | 779 |
Contrastive Visual Semantic Pretraining Magnifies the Semantics of Natural Language Representations | We examine the effects of contrastive visual semantic pretraining by comparing the geometry and semantic properties of contextualized English language representations formed by GPT-2 and CLIP, a zero-shot multimodal image classifier which adapts the GPT-2 architecture to encode image captions. We find that contrastive ... | Large-scale "natural language supervision" using image captions collected from the internet has enabled the first "zero-shot" artificial intelligence (AI) image classifiers, which allow users to create their own image classes using natural language, yet outperform supervised models on common language-and-image tasks Th... | in CWEs which outperform other static and contextualized word embeddings on wordlevel intrinsic evaluation tasks. CLIP word embeddings obtained in a "decontextualized" setting (wherein the model is given only the word with no other context) set new state of the art for a corpus-based method on the RG65 intrinsic evalua... | 1,612 | 821 | 1,612 |
Did You Mean...? Confidence-based Trade-offs in Semantic Parsing | We illustrate how a calibrated model can help balance common trade-offs in task-oriented parsing. In a simulated annotator-in-the-loop experiment, we show that well-calibrated confidence scores allow us to balance cost with annotator load, improving accuracy with a small number of interactions. We then examine how conf... | Task-oriented dialogue systems Recent work has focused on the calibration of semantic parsing models. Specifically, Stengel-Eskin and Van Durme (2022) benchmarked the calibration characteristics of a variety of semantic parsing models, finding some of them to be well-calibrated, especially on parsing for task-oriented ... | Our experiments in Section 4 involve a predictive model for human-in-the-loop coding: similar models have been integrated into IDEs, e.g. Datasets Our data is drawn from the SMCalFlow (Semantic Models We use MISO For token confidence estimation, we use the maximum probability across the output vocabulary at each timest... | 606 | 1,651 | 606 |
Regularized Structured Perceptron: A Case Study on Chinese Word Segmentation, POS Tagging and Parsing | Structured perceptron becomes popular for various NLP tasks such as tagging and parsing. Practical studies on NLP did not pay much attention to its regularization. In this paper, we study three simple but effective task-independent regularization methods: (1) one is to average weights of different trained models to red... | Structured perceptron is a linear classification algorithm. It is used for word segmentation The averaged perceptron or the voted perceptron Regularization is to improve the ability of generalization and avoid over-fitting for machine learning algorithms including online learning algorithms In this paper, we treat the ... | We treat the structured perceptron architecture as a multi-layer feed-forward neural network as in Figure The network of the structured perceptron has three layers. The input vector x and output vector y of the structured classification task are concatenated as the input layer. The hidden layer is the feature vector Φ(... | 440 | 1,538 | 440 |
Deterministic shift-reduce parsing for unification-based grammars by using default unification | Many parsing techniques including parameter estimation assume the use of a packed parse forest for efficient and accurate parsing. However, they have several inherent problems deriving from the restriction of locality in the packed parse forest. Deterministic parsing is one of solutions that can achieve simple and fast... | Over the last few decades, probabilistic unification-based grammar parsing has been investigated intensively. Previous studies Feature forests have been used successfully for probabilistic HPSG and CCG In this paper, we investigate shift-reduce parsing approach for unification-based grammars without the mechanisms of t... | A unification-based grammar is defined as a pair consisting of a set of lexical entries and a set of phrase-structure rules. The lexical entries express word-specific characteristics, while the phrase-structure rules describe constructions of constituents in parse trees. Both the phrasestructure rules and the lexical e... | 883 | 664 | 883 |
Discrete Latent Variable Representations for Low-Resource Text Classification | While much work on deep latent variable models of text uses continuous latent variables, discrete latent variables are interesting because they are more interpretable and typically more space efficient. We consider several approaches to learning discrete latent variable models for text in the case where exact marginali... | Deep generative models with latent variables have become a major focus of NLP research over the past several years. These models have been used both for generating text At the same time, deep generative models with discrete latent variables are attractive because the latents are arguably more interpretable, and because... | Our work builds on recent advances in discrete representation learning and its applications. In particular, we are inspired by recent success with VQ-VAEs outside NLP In addition to exploring the viability of VQ-VAEs for text representation learning, an important part of this paper is a systematic comparison between di... | 796 | 1,558 | 796 |
Joint Chinese Word Segmentation and Part-of-speech Tagging via Two-way Attentions of Auto-analyzed Knowledge | Chinese word segmentation (CWS) and partof-speech (POS) tagging are important fundamental tasks for Chinese language processing, where joint learning of them is an effective one-step solution for both tasks. Previous studies for joint CWS and POS tagging mainly follow the character-based tagging paradigm with introduci... | Chinese word segmentation (CWS) and part-ofspeech (POS) tagging are two fundamental and crucial tasks in natural language processing (NLP) for Chinese. The former one aims to find word boundaries in a sentence and the latter, on the top of segmentation results, assigns a POS tag to each word to indicate its syntactical... | The architecture of TWASP is illustrated in Figure To enhance the backbone paradigm, the proposed two-way attention module (as shown in the right part of Figure Auto-analyzed knowledge is demonstrated to be an effective type of resources to help NLP systems understand the texts be the sublists of S and K for x i . Here... | 1,509 | 2,226 | 1,509 |
Fidelity-Enriched Contrastive Search: Reconciling the Faithfulness-Diversity Trade-Off in Text Generation | In this paper, we address the hallucination problem commonly found in natural language generation tasks. Language models often generate fluent and convincing content but lack consistency with the provided source, resulting in potential inaccuracies. We propose a new decoding method called Fidelity-Enriched Contrastive ... | Language models (LMs) have achieved remarkable success in generating human-like text, fostering advancements across numerous Natural Language Processing (NLP) applications. Despite the fluent and seemingly convincing outputs produced by LMs, these models can occasionally generate content that is factually inconsistent ... | In this section, we present preliminary information on Contrastive Search To address shortcomings in existing decoding methods, Here, V k denotes a set of k candidate tokens with the top-k probability from the model's prediction distribution p θ (•|x 0:c+t ). The model confidence term represents the probability of the ... | 858 | 548 | 858 |
DOPA METER -A Tool Suite for Metrical Document Profiling and Aggregation | We present DOPA METER, a tool suite for the metrical investigation of written language, that provides diagnostic means for its division into discourse categories, such as registers, genres, and style. The quantitative basis of our system are 120 metrics covering a wide range of lexical, syntactic, and semantic features... | The way how we encode contents in natural language utterances gives rise to linguistic divisions into registers, genres, style levels, etc. (for a thorough distinction of these terms, see In this paper, we address a large variety of such behavioral aspects of language use from a metrical perspective. None of these metr... | The tool-based computational analysis of behavioral traits of language use can be divided into three branches of research: (1) readability checkers with language complexity measures incorporating mostly surface-level syntactic and lexicosemantic features of utterances, (2) stylometrics tools with strong emphasis on pow... | 543 | 540 | 543 |
Composing Finite State Transducers on GPUs | Weighted finite state transducers (FSTs) are frequently used in language processing to handle tasks such as part-of-speech tagging and speech recognition. There has been previous work using multiple CPU cores to accelerate finite state algorithms, but limited attention has been given to parallel graphics processing uni... | Finite-state transducers (FSTs) and their algorithms Composition is one of the most important operations on FSTs, because it allows complex FSTs to be built up from many simpler building blocks, but it is also one of the most expensive. Much work has been done on speeding up composition on a single CPU processor There ... | In this section, we introduce the notation that will be used throughout the paper for the composition task. A weighted FST is a tuple M = (Q, Σ, Γ, s, F, δ), where • Q is a finite set of states. • Σ is a finite input alphabet. • Γ is a finite output alphabet. • s ∈ Q is the start state. • F ⊆ Q are the accept states. N... | 660 | 650 | 660 |
Modeling Infant Word Segmentation | While many computational models have been created to explore how children might learn to segment words, the focus has largely been on achieving higher levels of performance and exploring cues suggested by artificial learning experiments. We propose a broader focus that includes designing models that display properties ... | The last fifteen years have seen an increased interest in the problem of how infants learn to segment a continuous stream of speech into words. Much of this work has been inspired by experiments with infants focusing on what capabilities infants have and which cues they attend to. While experimental work provides insig... | As this paper examines the intersection of infants' capabilities and computational modeling, we discuss work in both domains, beginning with experimental approaches to understanding how infants may perform the task of word segmentation. A potential account of how infants learn to identify words in fluent speech is that... | 631 | 1,781 | 631 |
SPT: Learning to Selectively Insert Prompts for Better Prompt Tuning | Prompt tuning prepends a soft prompt to the input embeddings or hidden states and only optimizes the prompt to adapt pretrained models (PTMs) to downstream tasks. The previous work manually selects prompt layers which are far from optimal and failed to exploit the potential of prompt tuning. In this work, we propose a ... | Increasingly large pre-trained models (PTMs) Prompt tuning IDPG In this paper, we first conduct a pilot experiment to show that simple modifications to the prompt inserting strategies in Our SPT framework considers a simple search space of whether to insert the generated instanceaware prompts into an intermediate layer... | A major research line of PETuning is the promptbased tuning that inserts some additional soft prompts into the embeddings or hidden states on specific layers of PTMs. Prompt tuning One important research line of PETuning is the adapter-based tuning Recently, there are work conducting automatic configurations of PETunin... | 982 | 1,188 | 982 |
ADVISER: A Dialog System Framework for Education & Research | In this paper, we present ADVISER 1 -an open source dialog system framework for education and research purposes. This system supports multi-domain task-oriented conversations in two languages. It additionally provides a flexible architecture in which modules can be arbitrarily combined or exchanged -allowing for easy s... | Dialog systems can be open-ended, e.g. small talk systems During the last years, several toolkits To address these shortcomings, we propose a multilingual multi-domain dialog system with two parallel goals: 1) to provide a highly flexible research framework not only for technique oriented developers but also for non-te... | During the last decade, several toolkits have been developed to facilitate the rapid implementation of goal-oriented dialog systems. RavenClaw Our approach is inline with InproTK OpenDial PyDial Our goal with this system is to provide both a highly modular research platform and an interdisciplinary educational tool. Us... | 644 | 422 | 644 |
KaggleDBQA: Realistic Evaluation of Text-to-SQL Parsers | The goal of database question answering is to enable natural language querying of real-life relational databases in diverse application domains. Recently, large-scale datasets such as Spider and WikiSQL facilitated novel modeling techniques for text-to-SQL parsing, improving zero-shot generalization to unseen databases... | Text-to-SQL parsing is a form of database question answering (DBQA) that answers a user's natural-language (NL) question by converting it into a SQL query over a given relational database. It can facilitate NL-based interfaces for arbitrary enduser applications, thereby removing the need for domain-specific UX or learn... | Text-to-SQL Semantic Parsing Semantic parsing has been studied extensively for decades In this paper, we propose a few-shot evaluation to inspire future research of practical text-to-SQL parsers. Like zero-shot, fewshot has access to many out-of-domain examples, but it also has access to a small number of indomain exam... | 1,087 | 2,698 | 1,087 |
Bilingual Lexicon Induction via Unsupervised Bitext Construction and Word Alignment | Bilingual lexicons map words in one language to their translations in another, and are typically induced by learning linear projections to align monolingual word embedding spaces. In this paper, we show it is possible to produce much higher quality lexicons with methods that combine (1) unsupervised bitext mining and (... | Bilingual lexicons map words in one language to their translations in another, and can be automatically induced by learning linear projections to align monolingual word embedding spaces We show that simply pipelining recent algorithms for unsupervised bitext mining These core contributions are established by systematic... | Bilingual lexicon induction (BLI). The task of BLI aims to induce a bilingual lexicon (i.e., word translation) from comparable monolingual corpora (e.g., Wikipedia in different languages). Following Word alignment. Word alignment is a fundamental problem in statistical machine translation, of which the goal is to align... | 1,034 | 751 | 1,034 |
Near-Negative Distinction: Giving a Second Life to Human Evaluation Datasets | Precisely assessing the progress in natural language generation (NLG) tasks is challenging, and human evaluation to establish a preference in a model's output over another is often necessary. However, human evaluation is usually costly, difficult to reproduce, and non-reusable. In this paper, we propose a new and simpl... | Pre-training of large language models has fueled recent progress in many natural language generation (NLG) tasks such as summarization The gold standard for NLG evaluation is manual expert annotation: it can be highly precise and fully customized to an NLG task, helping identify model limitations, and setting the direc... | We now detail the process of transforming preexisting human annotations into an NND dataset and show how to perform NND evaluation. A human annotation dataset D consists of (context, candidate) tuples that have been annotated typically with one or more labels from a discrete error categorization. Several properties are... | 1,219 | 2,241 | 1,219 |
DQ-BART: Efficient Sequence-to-Sequence Model via Joint Distillation and Quantization | Large-scale pre-trained sequence-to-sequence models like BART and T5 achieve state-ofthe-art performance on many generative NLP tasks. However, such models pose a great challenge in resource-constrained scenarios owing to their large memory requirements and high latency. To alleviate this issue, we propose to jointly d... | Pretrained sequence-to-sequence (seq2seq) models such as BART § Equal contribution. (2020) trained a BART model with 400M parameters, while The continual growth in model sizes leads to significant demand in both computation and memory resources during inference, and poses a huge challenge on deployment, especially in r... | In this section, we consider two directions for reducing the size of our generative language model: quantization ( §2.1) and distillation ( §2.2). We apply distillation-aware training ( §2.3) to train a quantized and distilled low-precision model as a student model to emulate the full-precision teacher model. Quantizat... | 1,064 | 2,035 | 1,064 |
OoMMix: Out-of-manifold Regularization in Contextual Embedding Space for Text Classification | Recent studies on neural networks with pretrained weights (i.e., BERT) have mainly focused on a low-dimensional subspace, where the embedding vectors computed from input words (or their contexts) are located. In this work, we propose a new approach, called OoMMix, to finding and regularizing the remainder of the space,... | Neural networks with a word embedding table have been the most popular approach to a wide range of NLP applications. The great success of transformer-based contextual embeddings as well as masked language models Along with outstanding performances of the pretrained weight, researchers have tried to reveal the underlyin... | In this section, we briefly review two approaches to regularizing over-parameterized network based on auxiliary tasks and auxiliary data. Regularization is an essential tool for good generalization capability of neural networks. One representative regularization approach relies on designing auxiliary tasks. Another app... | 1,149 | 2,311 | 1,149 |
Towards Generative Aspect-Based Sentiment Analysis * | Aspect-based sentiment analysis (ABSA) has received increasing attention recently. Most existing work tackles ABSA in a discriminative manner, designing various task-specific classification networks for the prediction. Despite their effectiveness, these methods ignore the rich label semantics in ABSA problems and requi... | Aspect-based sentiment analysis (ABSA), aiming at mining fine-grained opinion information towards specific aspects, has attracted increasing attention in recent years The main research line of ABSA focuses on the identification of those sentiment elements such as extracting the aspect term In general, most ABSA tasks a... | In this section, we describe the investigated ABSA tasks and the proposed two paradigms, namely, annotation-style and extraction-style modeling. Aspect Opinion Pair Extraction (AOPE) aims to extract aspect terms and their corresponding opinion terms as pairs Input: Salads were fantastic, our server was also very helpfu... | 970 | 2,128 | 970 |
Cross-language Sentence Selection via Data Augmentation and Rationale Training | This paper proposes an approach to crosslanguage sentence selection in a low-resource setting. It uses data augmentation and negative sampling techniques on noisy parallel sentence data to directly learn a cross-lingual embedding-based query relevance model. Results show that this approach performs as well as or better... | Sentence-level query relevance prediction is important for downstream tasks such as query-focused summarization and open-domain question answering; accurately pinpointing sentences containing information that is relevant to the query is critical to generating a responsive summary/answer (e.g., While we can use machine ... | Query-focused Sentence Selection Sentencelevel query relevance prediction is important for various downstream NLP tasks such as queryfocused summarization Sentence Selection A common approach to cross-language sentence selection is to use MT to first translate either the query or the sentence to the same language and t... | 777 | 1,920 | 777 |
Enhancing Extreme Multi-Label Text Classification: Addressing Challenges in Model, Data, and Evaluation | Extreme multi-label text classification is a prevalent task in industry, but it frequently encounters challenges in terms of machine learning perspectives, including model limitations, data scarcity, and time-consuming evaluation. This paper aims to mitigate these issues by introducing novel approaches. Firstly, we pro... | Extreme Multi-label Text Classification (XMTC) refers to the task of assigning to each document its most relevant labels from a taxonomy, where the number of labels could reach hundreds of thousands or millions However, the existing approaches often face inherent challenges pertaining to the model, data, and evaluation... | In the field of multi-label text classification, numerous studies have contributed to the development of effective models and techniques We introduce a label ranking model to replace the SciBERT-based model in our cooperative production. It comprises a Bi-Encoder model and a Cross-Encoder model. The Bi-Encoder model o... | 839 | 1,372 | 839 |
Beware of Model Collapse! Fast and Stable Test-time Adaptation for Robust Question Answering | Although pre-trained language models (PLM) have achieved great success in question answering (QA), their robustness is still insufficient to support their practical applications, especially in the face of distribution shifts. Recently, testtime adaptation (TTA) has shown great potential for solving this problem, which ... | Pre-trained language models (PLMs) have achieved great success on many NLP tasks To address this problem, researchers have proposed many approaches such as adversarial training To solve this problem, we take QA task as an example and investigate why TTA causes the model collapse. Our experiments indicate that the main ... | In this section, we begin by introducing extractive question answering and the application of TTA to enhance its robustness. Subsequently, we focus on Tent In extractive QA, the input of the model is a combination of a context and a question. The goal is to determine the start and end positions of the answer within the... | 1,292 | 1,711 | 1,292 |
Large Language Models in Machine Translation | This paper reports on the benefits of largescale statistical language modeling in machine translation. A distributed infrastructure is proposed which we use to train on up to 2 trillion tokens, resulting in language models having up to 300 billion n-grams. It is capable of providing smoothed probabilities for fast, sin... | Given a source-language (e.g., French) sentence f , the problem of machine translation is to automatically produce a target-language (e.g., English) translation ê. The mathematics of the problem were formalized by where {h m (e, f )} is a set of M feature functions and {λ m } a set of weights. One or more feature funct... | Traditionally, statistical language models have been designed to assign probabilities to strings of words (or tokens, which may include punctuation, etc.). Let w L 1 = (w 1 , . . . , w L ) denote a string of L tokens over a fixed vocabulary. An n-gram language model assigns a probability to w L 1 according to (2) where... | 539 | 1,575 | 539 |
Knowledge-Augmented Language Model Verification | Recent Language Models (LMs) have shown impressive capabilities in generating texts with the knowledge internalized in parameters. Yet, LMs often generate the factually incorrect responses to the given queries, since their knowledge may be inaccurate, incomplete, and outdated. To address this problem, previous works pr... | Recent Language Models (LMs) To mitigate hallucination of LMs, recent works have proposed to augment LMs with the knowledge retrieved from external knowledge sources (e.g., Wikipedia and Wikidata) In this work, we aim to overcome these suboptimalities of knowledge-augmented LMs. In other words, our goal is to verify wh... | Language Models Pre-trained Language Models (LMs) Knowledge-Augmented LMs Early works aim to incorporate knowledge from external knowledge sources (e.g., Wikipedia) into LMs, in order to enhance their performances on tasks that require factual knowledge, such as question answering. While such previous knowledge-augment... | 1,461 | 2,939 | 1,461 |
Evaluating and Improving Factuality in Multimodal Abstractive Summarization | Current metrics for evaluating factuality for abstractive document summarization have achieved high correlations with human judgment, but they do not account for the vision modality and thus are not adequate for visionand-language summarization. We propose CLIPBERTSCORE, a simple weighted combination of CLIPScore (Hess... | Multimodal abstractive summarization is the task of generating an abridged text that contains the most important information of the source inputs from various modalities. This challenging task builds upon the success of document summarization, where the input is only text documents. For document summarization, there ha... | Cut just the tip of the nails. Be sure you know where the quick is before you attempt to cut the nail … You should first cut just the tip of the nails ... 2. We create MUFAME, a meta-evaluation for factuality of multimodal summarization, and the WikiHowFact task to evaluate the quality of multimodal factuality metrics.... | 1,470 | 1,878 | 1,470 |
xPQA: Cross-Lingual Product Question Answering across 12 Languages | Product Question Answering (PQA) systems are key in e-commerce applications to provide responses to customers' questions as they shop for products. While existing work on PQA focuses mainly on English, in practice there is need to support multiple customer languages while leveraging product information available in Eng... | Product question answering (PQA) is a key technology in e-commerce applications. Given a question about a product, a PQA system searches the product webpage and provides an instant answer, so that customers do not need to traverse the page by themselves or seek help from humans | Figure 2017; To address this, we present xPQA, the first largescale dataset for cross-lingual PQA enabling non-English questions to be answered from English content. Most comprehensive product information is usually available in a majority language such as English. Therefore, searching for relevant information in Engli... | 615 | 278 | 615 |
A Case Study of Analysis of Construals in Language on Social Media Surrounding a Crisis Event | The events that took place at the Unite the Right rally held in Charlottesville, Virginia on August 11-12, 2017 caused intense reaction on social media from users across the political spectrum. We present a novel application of psycholinguistics -specifically, construal level theory -to analyze the language on social m... | Construal Level theory (CLT) To illustrate, consider the example of climate change. Research has shown that when people are primed to think about the topic of climate change using more concrete terms such as beetle and forest vs. more abstract terms (sea levels), they are more likely to engage with the topic of climate... | Construal Level Theory to Study Human Behavior: Construal level theory, first introduced by (2014) conducted a study regarding how psychological distance of thought would impact the positivity of reactions. They showed how distance from a scenario (having it happen to oneself versus to someone else) impacts one's react... | 564 | 1,746 | 564 |
Which Side are You on? Identifying Perspectives at the Document and Sentence Levels | In this paper we investigate a new problem of identifying the perspective from which a document is written. By perspective we mean a point of view, for example, from the perspective of Democrats or Republicans. Can computers learn to identify the perspective of a document? Not every sentence is written strongly from a ... | In this paper we investigate a new problem of automatically identifying the perspective from which a document is written. By perspective we mean a "subjective evaluation of relative significance, a point-of-view." (1) The inadvertent killing by Israeli forces of Palestinian civilians -usually in the course of shooting ... | Identifying the perspective from which a document is written is a subtask in the growing area of automatic opinion recognition and extraction. Subjective language is used to express opinions, emotions, and sentiments. So far, research in automatic opinion recognition has primarily addressed learning subjective language... | 789 | 3,101 | 789 |
Learning New Skills after Deployment: Improving open-domain internet-driven dialogue with human feedback | Frozen models trained to mimic static datasets can never improve their performance. Models that can employ internet-retrieval for up-to-date information and obtain feedback from humans during deployment provide the promise of both adapting to new information, and improving their performance. In this work we study how t... | Large language models employed as dialogue agents are primarily trained on human-written documents and human-human conversations collected from the web for pre-training | Input: What's happening in F1 these days? Response: F1 is a metric used for classification. In this work, we study learning from the feedback collected during deployment of models in human-model conversations. Such a setting has the opportunity to learn from within-distribution data, both in terms of the input contexts... | 1,012 | 168 | 1,012 |
Learn and Consolidate: Continual Adaptation for Zero-Shot and Multilingual Neural Machine Translation | Although existing multilingual neural machine translation (MNMT) models have demonstrated remarkable performance to handle multiple translation directions in a single model and achieved zero-shot translation between language pairs unseen in training, they still suffer from relatively poor translation qualities for some... | Existing multilingual neural machine translation (MNMT) models, such as mBART Fortunately, new parallel sentence pairs will continually emerge between high-resource lan- guages, e.g., German↔Chinese, which can be used to facilitate translation directions with poor performance through supervised learning In this scenari... | Zero-Shot Translation with MNMT Models MNMT models have demonstrated their ability to facilitate knowledge transfer across languages and enable zero-shot translations between language pairs that are not covered in training data For instance, Continual Learning for MNMT Some previous methods of continual learning attemp... | 1,025 | 2,424 | 1,025 |
A Simple Recipe towards Reducing Hallucination in Neural Surface Realisation | Recent neural language generation systems often hallucinate contents (i.e., producing irrelevant or contradicted facts), especially when trained on loosely corresponding pairs of the input structure and text. To mitigate this issue, we propose to integrate a language understanding module for data refinement with selftr... | Neural models for natural language generation (NLG) based on the encoder-decoder framework have become quite popular recently Given that similar issues have been less reported or noticed in the latest neural machine translation systems, we believe that the origin of the issue for neural NLG comes from the data side. Cu... | Our proposed framework consists of a neural natural language understanding (NLU) module with iterative data refinement to induce semantically equivalent MR-text pairs from a dataset containing a moderate level of noise. Formally, given a corpus with paired meaning representations and text descriptions {(R, X)} N i=1 , ... | 803 | 1,302 | 803 |
Seen to Unseen: Exploring Compositional Generalization of Multi-Attribute Controllable Dialogue Generation | Existing controllable dialogue generation work focuses on the single-attribute control and lacks generalization capability to out-of-distribution multiple attribute combinations. In this paper, we explore the compositional generalization for multi-attribute controllable dialogue generation where a model can learn from ... | Recently, large pre-trained language models (PLMs) like DialoGPT unseen multi-attribute combination seen multi-attribute combination (b) A-ACC Figure Although these methods have made some progress in CDG, most of them focus on singleattribute generation where there is only one attribute label like happiness in emotion ... | Controllable Dialogue Generation Currently, there have existed many studies on CDG As shown in Figure To better use the control signals, we design two types of prompts to elicit the attribute-related information from the PLM: Attribute-oriented Prompt We use the combination of controlled attribute values correspondin... | 878 | 2,198 | 878 |
Application-Agnostic Language Modeling for On-Device ASR | On-device automatic speech recognition systems face several challenges compared to server-based systems. They have to meet stricter constraints in terms of speed, disk size and memory while maintaining the same accuracy. Often they have to serve several applications with different distributions at once, such as communi... | On-device Automatic Speech Recognition (ASR) is subject to several constraints: it should return accurate results in a reasonable time frame without consuming too much memory and disk space. State-of-the-art research often is accuracy focused, while resource-constrained applications also need to take care of performanc... | We start by discussing related work on modeling several domains/tasks at once. Many pattern recognition tasks are imbalanced since data from different categories do not occur at the same frequency. Therefore, the less frequent categories are not well represented in the training data The choice of architecture for langu... | 943 | 2,635 | 943 |
Improving Passage Retrieval with Zero-Shot Question Generation | We propose a simple and effective re-ranking method for improving passage retrieval in open question answering. The re-ranker re-scores retrieved passages with a zero-shot question generation model, which uses a pre-trained language model to compute the probability of the input question conditioned on a retrieved passa... | Text retrieval is a core sub-task in many NLP problems, for example, open-domain question answering where a document must be retrieved and then read to answer an input query. Queries and documents are typically embedded in a shared representation space to enable efficient search, before using a task-specific model to p... | An open-domain QA system consists of a retriever and a reader component. The reader attends to the retrieved passages to produce a final answer to the question. We use the Fusion-in-Decoder (FiD; Izacard and Grave (2021b)) model as the reader. In FiD, each retrieved passage is concatenated with the question and is then... | 1,050 | 1,583 | 1,050 |
Tchebycheff Procedure for Multi-task Text Classification | Multi-task Learning methods have achieved significant progress in text classification. However, existing methods assume that multi-task text classification problems are convex multiobjective optimization problems, which is unrealistic in real-world applications. To address this issue, this paper presents a novel Tcheby... | Multi-task Learning (MTL) aims to learn multiple related tasks simultaneously, and obtain better performance than learning each task independently by setting inductive bias across tasks. Existing MTL methods for text classification, usually set up the inductive bias across tasks by designing a parameterized hypothesis ... | The family of Pareto optimality methods, including L 1 -metric methods (weighted sum methods) To handle the non-convex case, MGDA leverages the Karush-Kuhn-Tucker conditions and provides Pareto stationary points as solutions. However, the solutions are not sufficient to be Pareto optimal. A novel MTL method, which can ... | 522 | 1,382 | 522 |
Few-Shot Dialogue Summarization via Skeleton-Assisted Prompt Transfer in Prompt Tuning | In real-world scenarios, labeled samples for dialogue summarization are usually limited (i.e., few-shot) due to high annotation costs for high-quality dialogue summaries. To efficiently learn from few-shot samples, previous works have utilized massive annotated data from other downstream tasks and then performed prompt... | Automatic text summarization | The user asks for the address, postcode and phone number of Oriental House. The restaurant is in the east and the food is expensive. Few-Shot Figure days In existing works, one common way to tackle the data scarcity problem is to perform transfer learning by leveraging off-the-shelf out-of-domain or out-of-task supervi... | 1,396 | 28 | 1,396 |
Saying No is An Art: Contextualized Fallback Responses for Unanswerable Dialogue Queries | Despite end-to-end neural systems making significant progress in the last decade for taskoriented as well as chit-chat based dialogue systems, most dialogue systems rely on hybrid approaches which use a combination of rulebased, retrieval and generative approaches for generating a set of ranked responses. Such dialogue... | In order to cater to the diversity of questions spanning across various domains, dialogue systems generally follow a hybrid architecture wherein an ensemble of individual response subsystems One approach to acknowledge such queries is to have a fallback mechanism with responses like "I don't know the answer to this que... | Improving the coverage to address out-of-domain queries is not a new problem in designing dialogue systems. The most popular approach has been via presenting the user with chit-chat responses. Other systems such as Blender We describe two approaches to generate such contextual don't-know responses. Inspired by previous... | 1,228 | 1,235 | 1,228 |
Normalizing Mutual Information for Robust Adaptive Training for Translation | Despite the success of neural machine translation models, tensions between fluency of optimizing target language modeling and sourcefaithfulness remain as challenges. Previously, Conditional Bilingual Mutual Information (CBMI), a scoring metric for the importance of target sentences and tokens, was proposed to encourag... | Neural machine translation (NMT) models have achieved remarkable performance since Conditional Bilingual Mutual Information (CBMI), a metric for target tokens and sentences computed as the log quotient of the translation and the target-side language model probability, was While our proposed sentence-level NPMI assigns ... | 6 Related work Given two random variables X and Y , the pointwise mutual information (PMI) between the observations x and y is which is not bounded below and has an upper bound oflog p(x, y). Token-level adaptive training, inspired by earlier approaches to fighting class imbalance problem in classification tasks, aims ... | 1,106 | 1,380 | 1,106 |
COLT5: Faster Long-Range Transformers with Conditional Computation | Many natural language processing tasks benefit from long inputs, but processing long documents with Transformers is expensive --not only due to quadratic attention complexity but also from applying feedforward and projection layers to every token. However, not all tokens are equally important, especially for longer doc... | Many natural language processing tasks, such as summarization Over the past few years, many "efficient Transformer" approaches have been proposed that reduce the cost of the attention mechanism over long inputs This paper presents COLT5 (Conditional LongT5), a new family of models that, building on top of LONGT5 In par... | Transformer FLOPs COLT5 follows an extensive line of work in attempting to reduce the computational cost of Transformer models, particularly over long inputs. The computational burden of Transformer models has several distinct elements, and different approaches focus on reducing the cost of different components. For th... | 810 | 1,174 | 810 |
Automated Chess Commentator Powered by Neural Chess Engine | In this paper, we explore a new approach for automated chess commentary generation, which aims to generate chess commentary texts in different categories (e.g., description, comparison, planning, etc.). We introduce a neural chess engine into text generation models to help with encoding boards, predicting moves, and an... | With games exploding in popularity, the demand for Natural Language Generation (NLG) applications for games is growing rapidly. Related researches about generating real-time game reports It is common knowledge that professional game commentators are usually game players. And expert players can usually provide more thor... | The most relevant work is Data-to-text generation is a popular track in NLG researches. Recent researches are mainly about generating from structured data to biography The overview of our approach is shown in Figure In Figure Description Model: Descriptions about the current move intuitively depend on the move itself. ... | 619 | 1,201 | 619 |
Simultaneous Translation Policies: From Fixed to Adaptive | Adaptive policies are better than fixed policies for simultaneous translation, since they can flexibly balance the tradeoff between translation quality and latency based on the current context information. But previous methods on obtaining adaptive policies either rely on complicated training process, or underperform s... | Simultaneous translation (ST) aims to provide good translation quality while keeping the latency of translation process as low as possible. This is very important for the scenarios that require simultaneity, such as international summits and negotiations. For this, human interpreters usually start translation before th... | Full-sentence translation. Neural machine translation (NMT) model usually consists of two components: an encoder, which encodes the source sentence x = (x 1 , . . . , x m ) into a sequence of hidden states, and a decoder, which sequentially predicts target tokens conditioned on those hidden states and previous predicti... | 773 | 1,317 | 773 |
hyperdoc2vec: Distributed Representations of Hypertext Documents | Hypertext documents, such as web pages and academic papers, are of great importance in delivering information in our daily life. Although being effective on plain documents, conventional text embedding methods suffer from information loss if directly adapted to hyper-documents. In this paper, we propose a general embed... | The ubiquitous World Wide Web has boosted research interests on hypertext documents, e.g., personal webpages To model hypertext documents, various efforts Conventional attempts on utilizing embedding techniques in hyper-doc-related tasks generally fall into two types. The first type • What information should hyper-doc ... | Network representation learning is a related topic to ours since a collection of hyper-docs resemble a network. To embed nodes in a network, Document embedding for classification is another focused area to apply document embeddings. Le and Citation recommendation is a direct downstream task to evaluate embeddings learn... | 787 | 1,616 | 787 |
ODE Transformer: An Ordinary Differential Equation-Inspired Model for Sequence Generation | Residual networks are an Euler discretization of solutions to Ordinary Differential Equations (ODE). This paper explores a deeper relationship between Transformer and numerical ODE methods. We first show that a residual block of layers in Transformer can be described as a higher-order solution to ODE. Inspired by this,... | Residual networks have been used with a great success as a standard method of easing information flow in multi-layer neural models where F (•, •) is the function of the layer and θ t is its parameter. Interestingly, recent work in machine learning (2) * Corresponding author. where y(t) and θ(t) are continuous with resp... | We start with a description of Transformer, followed by its relationship with ODEs. We choose Transformer for our discussion and experiments because it is one of the state-of-the-art models in recent sentence generation tasks. Transformer is an example of the encoder-decoder paradigm where LN(•) is the layer normalizat... | 927 | 1,558 | 927 |
Exploring Distributional Shifts in Large Language Models for Code Analysis | We systematically study how three large language models with code capabilities -CodeT5, Codex, and ChatGPT -generalize to out-ofdomain data. We consider two fundamental applications -code summarization, and code generation. We split data into domains following its natural boundaries -by an organization, by a project, a... | Since the late 2000s, researchers have been reporting poor generalization of statistical learning models to new software systems However, the challenges of distribution shifts stemming from the hierarchical nature of software data, as depicted in Figure Figure Next, we explore ways to improve the out-ofdomain generaliz... | The shifts in underlying semantics between the training and evaluation data can be one of the most impacting factors for deteriorating performance at test time. Prior work in code analysis has mainly focused on cross-project shifts, i.e. training and evaluating the model on disjunct sets of code projects. Additionally,... | 1,126 | 1,623 | 1,126 |
Weakly Supervised Semantic Parsing with Abstract Examples | Training semantic parsers from weak supervision (denotations) rather than strong supervision (programs) complicates training in two ways. First, a large search space of potential programs needs to be explored at training time to find a correct program. Second, spurious programs that accidentally lead to a correct denot... | The goal of semantic parsing is to map language utterances to executable programs. Early work on statistical learning of semantic parsers utilized * Authors equally contributed to this work. | k :[[{y loc: ..., color: 'Black', type: 'square', x loc: ... size: 20}, .. x :There is a small yellow item not touching any wall y :True z :Exist(Filter(ALL ITEMS, λx. Figure supervised learning, where training examples included pairs of language utterances and programs Training semantic parsers from denotations rather... | 1,239 | 190 | 1,239 |
Arabic Dialect Identification with a Few Labeled Examples Using Generative Adversarial Networks | Given the challenges and complexities introduced while dealing with Dialect Arabic (DA) variations, Transformer based models, e.g., BERT, outperformed other models in dealing with the DA identification task. However, to fine-tune these models, a large corpus is required. Getting a large number high quality labeled exam... | While Arabic is the first language of most of the Middle East and North Africa (MENA) region, different countries have different dialects of Arabic. These Dialect Arabic (DA) forms are all different from the Modern Standard Arabic (MSA). MSA is used in formal writing and speaking situations, like academia and media. In... | The main challenge in Arabic Dialect Identification is the rarity of high-quality labeled datasets that represent all Arabic dialects. Recently, some datasets were introduced. However, most of them have limitations as will be shown in the next paragraphs. The Arabic Online Commentary AOC Dialect Identification shared t... | 970 | 2,123 | 970 |
KEBAP: Korean Error Explainable Benchmark Dataset for ASR and Post-processing | Automatic Speech Recognition (ASR) systems are instrumental across various applications, with their performance being critically tied to user satisfaction. Conventional evaluation metrics for ASR systems produce a singular aggregate score, which is insufficient for understanding specific system vulnerabilities. Therefo... | Automatic speech recognition (ASR) is a task that recognizes speech and converts it into text, and it is getting more and more attention with the development of voice interface applications and devices such as Alexa, Siri, and Cortana if the ASR model accurately recognizes the input voice, the user's readability may de... | In the real-world scenario, mitigating the tradeoff between recognition accuracy and user readability is crucial. To address this, we propose KE-BAP, emphasizing the importance of considering both aspects. A detailed explanation is as follows. Firstly, in real-world speech recognition, it is essential to consider the a... | 1,471 | 2,076 | 1,471 |
Scheduled Multi-task Learning for Neural Chat Translation | Neural Chat Translation (NCT) aims to translate conversational text into different languages. Existing methods mainly focus on modeling the bilingual dialogue characteristics (e.g., coherence) to improve chat translation via multi-task learning on small-scale chat translation data. Although the NCT models have achieved... | A cross-lingual conversation involves speakers in different languages (e.g., one speaking in Chinese and another in English), where a chat translator can be applied to help them communicate in their native languages. The chat translator bilaterally converts the language of bilingual conversational text, e.g. from Chine... | In a bilingual conversation, we assume the two speakers have alternately given utterances in different languages for u turns, resulting in X 1 , X 2 , X 3 , ..., X u and Y 1 , Y 2 , Y 3 , ..., Y u on the source and target sides, respectively. Among these utterances, X 1 , X 3 , X 5 , ..., X u are originally spoken and ... | 1,231 | 1,375 | 1,231 |
Adversarial Stylometry in the Wild: Transferable Lexical Substitution Attacks on Author Profiling | Written language contains stylistic cues that can be exploited to automatically infer a variety of potentially sensitive author information. Adversarial stylometry intends to attack such models by rewriting an author's text. Our research proposes several components to facilitate deployment of these adversarial attacks ... | The widespread use of machine learning on consumer devices and its application to their data has sparked investigation of security and privacy researchers alike in correctly handling sensitive information Privacy-preserving defenses against such inferences can be found in the field of adversarial Adversarial attacks on... | Stylometry, the study of (predominantly) writing style, dates back several decades It is perhaps for this reason that most obfuscation work uses heuristically-driven, controlled changes such as splitting or merging words or sentences, removing stop words, changing spelling, punctuation, or casing (see e.g., Our work do... | 807 | 2,766 | 807 |
Learning to Generate Task-Specific Adapters from Task Description | Pre-trained text-to-text transformers such as BART have achieved impressive performance across a range of NLP tasks. Recent study further shows that they can learn to generalize to novel tasks, by including task descriptions as part of the source sequence and training the model with (source, target) examples. At test t... | Pre-trained text-to-text models unseen tasks with the source sequence containing new task descriptions. While this initial attempt shows positive results, there are two potential limitations for the direct finetuning approach. (1) Predictions can be sensitive to the task descriptions (or "prompts") that are heuristical... | We study the problem of learning from task description For instance, in the ZEST dataset Our work is built on adapters Overview. Fig. (1) A main network, which is a pre-trained text-totext model. We instantiate the main network with BART-Base/Large Hypernetwork. The hypernetwork consists of an encoder and multiple deco... | 1,152 | 640 | 1,152 |
Entity-Focused Dense Passage Retrieval for Outside-Knowledge Visual Question Answering | Most Outside-Knowledge Visual Question Answering (OK-VQA) systems employ a twostage framework that first retrieves external knowledge given the visual question and then predicts the answer based on the retrieved content. However, the retrieved knowledge is often inadequate. Retrievals are frequently too general and fai... | Passage retrieval under a multi-modal setting is a critical prerequisite for applications such as outsideknowledge visual question answering (OK-VQA) In this work, we investigate two main drawbacks of recent dense retrievers First, as most retrieval models encode the query and passages as a whole, they fail to explicit... | Visual Question Answering (VQA) has witnessed remarkable progress over the past few years, in terms of both the scope of the questions Sparse Retrieval: Before the recent proliferation of transformer-based dense passage retrieval models Motivated by the trend toward dense retrievers, previous work has also applied them... | 1,010 | 1,278 | 1,010 |
EnsLM: Ensemble Language Model for Data Diversity by Semantic Clustering | Natural language processing often faces the problem of data diversity such as different domains, themes, styles and so on. Therefore, a single language model (LM) is insufficient to learn all knowledge from diverse samples. To solve this problem, we firstly propose an autoencoding topic model with mixture prior (mATM) ... | It is common knowledge in modern natural language processing (NLP) that natural language varies greatly across domains, themes, styles, genres and many other linguistic nuances (Van der Data selection is a commonly used strategy to handle diversity in data (Moore and Lewis, 2010; Inspired by their works and to move bey... | For NLP, topic modeling (TM) 3 Autoencoding topic model with mixture prior We firstly describe one of the most popular topic models, latent Dirichlet allocation (LDA) For a document containing D words as where φ k is a probability distribution over the vocabulary, LDA defines the generative process of w in Algorithm 1,... | 1,001 | 810 | 1,001 |
Temporal Scoping of Relational Facts based on Wikipedia Data | extraction from text has focused on named-entity recognition, entity linking, and relation extraction. Less attention has been paid given to extracting the temporal scope for relations between named entities; for example, the relation president-Of(John F. Kennedy, USA) is true only in the time-frame (January 20, 1961 ... | Previous work on relation extraction In this paper, we describe TSRF, a system for temporal scoping of relational facts. For every relation type, TSRF uses distant supervision from Wikipedia infobox tuples to learn a language model consisting of patterns of entity types, categories, and word n-grams. Then it uses this ... | To our knowledge, there are only a small number of systems that have tackled the temporal scoping of relations task. YAGO The TempEval task The current state-of-the-art systems for TSF have been the RPI-Blender system by 3 The Temporal Slot Filling Task The input format for a TSF system as instantiated for the relation... | 989 | 1,218 | 989 |
SIB-200: A Simple, Inclusive, and Big Evaluation Dataset for Topic Classification in 200+ Languages and Dialects | Despite the progress in building multilingual language models, evaluation is often limited to a few languages with available datasets which excludes a large number of low-resource languages. In this paper, we create SIB-200a large-scale open-sourced benchmark dataset for topic classification in 205 languages and dialec... | In the last few years, developing massively multilingual Pre-trained Language Models (PLMs) to scale to several written languages is an active area of research-e.g. covering 100 languages While there is evidence from previous works that languages not covered during pre-training often lead to lower performance, such ana... | We introduce SIB-200-a Simple Inclusive and Big topic classification dataset for over 200 languages and dialects. We leveraged the multi-way parallel Flores-200 dataset annotate (2,009 rather than 562 instances We recruited four annotators who are native speakers of English to label 2,009 sentences obtained from the DE... | 1,499 | 1,975 | 1,499 |
Selecting Backtranslated Data from Multiple Sources for Improved Neural Machine Translation | Machine translation (MT) has benefited from using synthetic training data originating from translating monolingual corpora, a technique known as backtranslation. Combining backtranslated data from different sources has led to better results than when using such data in isolation. In this work we analyse the impact that... | The use of supplementary backtranslated text has led to improved results in several tasks such as automatic post-editing (Junczys-Dowmunt and While In this work we conduct a systematic study of the effects of backtranslated data from different sources, as well as how to optimally select subsets of this data taking into... | One of the first papers comparing the performance of different systems for backtranslation was More recently In this work we extend these ideas by combining backtranslated data from RBMT, PB-SMT, NMT (LSTM) and While the most common approach to assessing the translation capabilities of a MT system is via evaluation sco... | 1,166 | 1,095 | 1,166 |
HAT: Hardware-Aware Transformers for Efficient Natural Language Processing | Transformers are ubiquitous in Natural Language Processing (NLP) tasks, but they are difficult to be deployed on hardware due to the intensive computation. To enable low-latency inference on resource-constrained hardware platforms, we propose to design Hardware-Aware Transformers (HAT) with neural architecture search. ... | Transformer Nevertheless, it is challenging to deploy Transformers on mobile devices due to the high computation cost. For instance, in order to translate a sentence with only 30 words, a Transformer-Big model needs to execute 13G FLOPs and takes 20 seconds on a Raspberry Pi. Such long latency will hurt the user experi... | An overview of the HAT framework is shown in Figure We construct a large design space by breaking two conventions in the Transformer design: (1) All decoder layers only attend to the last encoder layer; (2) All the layers are identical. Arbitrary Encoder-Decoder Attention. Different encoder layers extract features on d... | 1,152 | 2,004 | 1,152 |
Math Word Problem Solving by Generating Linguistic Variants of Problem Statements | The art of mathematical reasoning stands as a fundamental pillar of intellectual progress and is a central catalyst in cultivating human ingenuity. Researchers have recently published a plethora of works centered around the task of solving Math Word Problems (MWP) -a crucial stride towards general AI. These existing mo... | Math word problem solving is a long-standing research problem in Artificial General Intelligence (AGI) and a lot of studies about this topic, from both industry and academia, have been published recently. A typical Math Word Problem (MWP) takes the form of a written narrative that articulates a problem scenario and pos... | The dawn of research on MWP solving was in the Currently, the landscape of Deep learning models for the MWP solving task is primarily comprised of five distinct paradigms, SEQ2SEQbased, SEQ2TREE-based, GRAPH2TREE-based, complex relation extraction-based, and Large Language Model (LLM) prompt-based approaches, each of w... | 1,430 | 1,380 | 1,430 |
Integrated Learning of Dialog Strategies and Semantic Parsing | Natural language understanding and dialog management are two integral components of interactive dialog systems. Previous research has used machine learning techniques to individually optimize these components, with different forms of direct and indirect supervision. We present an approach to integrate the learning of b... | Natural language understanding and dialog management are two integral components of a dialog system. Current research typically deals with optimizing only one of these components. We present an approach to integrate the learning of both a dialog strategy using reinforcement learning, and a semantic parser for robust na... | Prior work has used dialog to facilitate robot task learning, e.g. There has been considerable work in semantic parsing using both direct supervision in the form of annotated meaning representations There has also been considerable work in goaldirected dialog systems in domains such as information provision More recent... | 680 | 2,536 | 680 |
PANACEA: An Automated Misinformation Detection System on COVID-19 | In this demo, we introduce a web-based misinformation detection system PANACEA on COVID-19 related claims, which has two modules, fact-checking and rumour detection. Our fact-checking module, which is supported by novel natural language inference methods with a self-attention network, outperforms state-ofthe-art approa... | The dangers of misinformation have become even more apparent to the general public during the COVID-19 pandemic. Following false treatment information has led to a high number of deaths and hospitalisations In this work, we focus on automating misinformation detection using information from credible sources as well as ... | The following datasets are used in the project: Knowledge Database This is used for factchecking, and includes COVID-19 related documents from selected reliable sources PANACEA Dataset In order to fine-tune our model, we constructed a new COVID-19 related propagation tree dataset for rumour detection. Similar previous ... | 783 | 1,936 | 783 |
CBNU System for SIGMORPHON 2019 Shared Task 2: a Pipeline Model | In this paper we describe our system for morphological analysis and lemmatization in context, using a transformer-based sequence to sequence model and a biaffine attention based BiLSTM model. First, a lemma is produced for a given word, and then both the lemma and the given word are used for morphological analysis. We ... | In this paper we present our neural network architecture that we have used for the SIGMORPHON 2019 shared task 2 Hence, morphological analysis/tagging is a classification task for an input sequence. | There are two tasks in SIGMORPHON 2019 and we chose task 2. The idea of the task is simple: the input is a sentence made of words and the output is a lemma and morphosyntactic description (MSD) for each word. Table The dataset consists of initial 98 datasets of more than 60 distinct languages, and additional nine surpr... | 608 | 198 | 608 |
From Discourse to Narrative: Knowledge Projection for Event Relation Extraction | Current event-centric knowledge graphs highly rely on explicit connectives to mine relations between events. Unfortunately, due to the sparsity of connectives, these methods severely undermine the coverage of EventKGs. The lack of high-quality labelled corpora further exacerbates that problem. In this paper, we propose... | Event-centric knowledge graphs (EventKGs) model the narratives of the world by representing events and identifying relations between them, which are critical for machine understanding and can benefit many downstream tasks, such as question answering Recently, semi-automatically constructing Even-tKGs have gained much a... | Event Relation Extraction (ERE). connectives, while implicit relations lack these surface cues. To resolve the implicit discourse relation recognition (IDRR) task, researchers construct high-quality labelled datasets Associations between Discourse and Narrative. Recent NLP studies have proved that discourse and narrati... | 949 | 2,467 | 949 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.