---
language:
  - en
license: cc-by-sa-4.0
task_categories:
  - token-classification
  - text-classification
task_ids:
  - named-entity-recognition
  - part-of-speech
  - parsing
tags:
  - ner
  - pos
  - dependency-parsing
  - dialog-act
  - multi-task
  - conllu
  - spacy
  - gpt-validated
size_categories:
  - 100K<n<1M
---

# kniv-corpus-en

Multi-domain NLP corpus for English with annotations for NER, POS tagging, dependency parsing, and sentence classification (dialog act). Built for training [kniv](https://github.com/rustic-ai/kniv-nlp-models) multi-task NLP models for the [uniko](https://github.com/rustic-ai/uniko) cognitive memory system.

## Dataset Details

| | |
|---|---|
| **Format** | CoNLL-U with NER and CLS annotations in metadata/MISC fields |
| **Sentences** | 785,722 |
| **Annotation** | spaCy `en_core_web_trf` + GPT-5.4 validation corrections |
| **CLS labels** | GPT-5.4-nano (9 dialog act types) |
| **License** | CC BY-SA 4.0 |

### Splits

| Split | Sentences |
|-------|-----------|
| train | 628,577 |
| dev | 78,572 |
| test | 78,573 |

### Domains

| Domain | Sentences | Sources | License |
|--------|-----------|---------|---------|
| **Conversation** | 218,343 | Taskmaster 1/2/3, OASST1, MultiWOZ 2.2, Glaive Function Calling, Discord Dialogues | CC BY-4.0 / Apache 2.0 / MIT |
| **Business** | 197,891 | SEC EDGAR 10-K, Enron emails, OpenStax textbooks, Odoo docs, Wikipedia, CUAD contracts, OpenAlex abstracts | Public domain / CC BY-4.0 / CC BY-SA 3.0 / ODC-BY |
| **Technical** | 193,032 | Wikipedia (CS/engineering), Python documentation | CC BY-SA 3.0 / PSF License |
| **Encyclopedic** | 66,490 | Wikipedia (science, history, geography, arts) | CC BY-SA 3.0 |
| **News** | 58,026 | Wikinews, Wikipedia (journalism, politics) | CC BY 2.5 / CC BY-SA 3.0 |
| **Narrative** | 51,940 | Project Gutenberg (classic fiction) | Public domain |

### Annotation Pipeline

1. **Collect** — raw text from open sources (APIs, HuggingFace datasets, git repos)
2. **Preprocess** — sentence splitting, cleaning, quality filtering, deduplication
3. **Annotate** — spaCy `en_core_web_trf` produces POS, NER, dependency parse per sentence
4. **Validate** — GPT-5.4-nano checks annotations, returns corrections for POS/NER/dep errors
5. **Classify** — GPT-5.4-nano assigns dialog act labels (9 types)
6. **Export** — corrections applied, CLS labels embedded, split into train/dev/test

### NER Entity Types (18)

BIO-tagged in the MISC column:

| Type | Description | Example |
|------|-------------|---------|
| PERSON | People | *Caroline*, *Dr. Smith* |
| ORG | Organizations | *Apple*, *United Nations* |
| GPE | Geopolitical entities | *France*, *New York* |
| LOC | Locations | *Mount Everest*, *Pacific Ocean* |
| DATE | Dates and periods | *January 2024*, *last quarter* |
| TIME | Times | *3pm*, *two hours* |
| MONEY | Monetary values | *$394B*, *50 million euros* |
| PERCENT | Percentages | *15.3%*, *a third* |
| QUANTITY | Measurements | *100 kilometers*, *5 pounds* |
| ORDINAL | Ordinal numbers | *first*, *3rd* |
| CARDINAL | Cardinal numbers | *three*, *42* |
| NORP | Nationalities/groups | *British*, *Republican* |
| FAC | Facilities | *the White House*, *Highway 101* |
| PRODUCT | Products | *iPhone*, *Boeing 747* |
| EVENT | Events | *World War II*, *Olympics* |
| WORK_OF_ART | Works of art | *Hamlet*, *The Starry Night* |
| LAW | Laws | *the First Amendment* |
| LANGUAGE | Languages | *English*, *Mandarin* |

### CLS Dialog Act Labels (9)

Embedded as `# cls = <label>` comment in each sentence:

| Label | Action in uniko | Example |
|-------|----------------|---------|
| `inform` | Extract observation | *"Revenue grew 15% last quarter"* |
| `correction` | Flag knowledge for update | *"No, it was actually Tuesday"* |
| `agreement` | Reinforce observation | *"Yes, exactly"* |
| `question` | Record knowledge gap | *"Where is the report?"* |
| `plan_commit` | Link to Goal/Task | *"I'll send it tomorrow"* |
| `request` | Create action node | *"Show me the data"* |
| `feedback` | Skip extraction | *"OK"*, *"Got it"* |
| `social` | Skip extraction | *"Hi"*, *"Thanks"* |
| `filler` | Skip entirely | *"Um"*, *"Well"* |

### CoNLL-U Format

Each sentence has metadata comments followed by token rows:

```
# sent_id = business-000042
# text = Apple reported $394B revenue in Q4.
# cls = inform
1    Apple      apple    PROPN    NNP    _    2    nsubj     _    NER=B-ORG
2    reported   report   VERB     VBD    _    0    root      _    _
3    $394B      $394b    NUM      CD     _    2    obj       _    NER=B-MONEY
4    revenue    revenue  NOUN     NN     _    3    appos     _    _
5    in         in       ADP      IN     _    6    case      _    _
6    Q4         q4       PROPN    NNP    _    2    obl       _    NER=B-DATE
7    .          .        PUNCT    .      _    2    punct     _    _
```

Columns: ID, FORM, LEMMA, UPOS, XPOS, FEATS, HEAD, DEPREL, DEPS, MISC

## Usage

```python
import conllu
from huggingface_hub import hf_hub_download

# Download a split
path = hf_hub_download("dragonscale-ai/kniv-corpus-en", "train.conllu", repo_type="dataset")

with open(path) as f:
    sentences = conllu.parse(f.read())

for sent in sentences[:3]:
    # Metadata
    print(f"Text: {sent.metadata['text']}")
    print(f"CLS:  {sent.metadata.get('cls', 'N/A')}")

    # Tokens
    for token in sent:
        ner = token.get("misc", {})
        ner_tag = ner.get("NER", "O") if isinstance(ner, dict) else "O"
        print(f"  {token['form']:20s} POS={token['upos']:6s} DEP={token['deprel']:12s} NER={ner_tag}")
    print()
```

## Source

- **Training code:** [rustic-ai/kniv-nlp-models](https://github.com/rustic-ai/kniv-nlp-models)
- **Trained models:** [dragonscale-ai on HuggingFace](https://huggingface.co/dragonscale-ai)

## Citation

```bibtex
@misc{kniv-corpus-en,
  title={kniv-corpus-en: Multi-domain NLP Corpus for English},
  author={Dragonscale AI},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/dragonscale-ai/kniv-corpus-en}
}
```

## License

CC BY-SA 4.0. Individual domain source licenses are listed in the domains table above and in the [source repository](https://github.com/rustic-ai/kniv-nlp-models/tree/main/corpus/domains).