--- license: apache-2.0 language: - zh - en widget: - text: TurnSense three-class speech turn detection demo output: url: image/PR_new.mp4 ---
Baiji Team Logo
# TurnSense ### ๐ŸŽฏ Lightweight ยท Accurate ยท Three-Class โ€” Redefining Speech Turn Detection
47M Parameters ๏ฝœ CPU Latency ~55ms ๏ฝœ F1 up to 96.35% ๏ฝœ Invalid Utterance Filtering

[![GitHub](https://img.shields.io/badge/GitHub-brgroup--Team/TurnSense-181717?style=for-the-badge&logo=github)](https://github.com/Bairong-Xdynamics/TurnSense) [![Hugging Face](https://img.shields.io/badge/๐Ÿค—_Hugging_Face-brgroup--Team-yellow?style=for-the-badge)](https://huggingface.co/brgroup/TurnSense) [![License](https://img.shields.io/badge/License-Apache_2.0-blue?style=for-the-badge)](./LICENSE) [![PRs Welcome](https://img.shields.io/badge/PRs-Welcome-brightgreen?style=for-the-badge)](https://github.com/Bairong-Xdynamics/TurnSense)

**Language**: **English** | [ไธญๆ–‡](./README_zh.md)
> **โญ If TurnSense is useful to you, please give us a Star!** This helps us continue improving the model and documentation.
## ๐Ÿ“– Table of Contents - [News](#-news) - [Why TurnSense](#-why-turnsense) - [Introduction](#-introduction) - [Core Features](#-core-features) - [Model Size Comparison](#-model-size-comparison) - [Benchmark Results](#-benchmark-results) - [Quick Start](#-quick-start) - [Evaluation Guide](#-evaluation-guide) - [Citation](#-citation) - [Questions and Contact](#-questions-and-contact) - [License](#-license)
---
## ๐Ÿ“ฐ News - **2026.05.22**: Released **TurnSense 1.1**, an English-enhanced version focused on improving `complete / incomplete` semantic completeness detection in English scenarios. It is suitable for Chinese-English mixed dialogue scenarios. The model is available on Hugging Face: [brgroup/TurnSense](https://huggingface.co/brgroup/TurnSense).
---
## ๐Ÿ† Why TurnSense
| Dimension | TurnSense Performance | | :---: | :---: | | ๐ŸŽฏ **Accuracy** | F1 **96.35%** on `easyturn_real_test_ZH` โ€” best among comparable models | | โšก **Inference Latency** | CPU p50 โ‰ˆ **54.65ms** โ€” suitable for real-time interaction | | ๐Ÿ“ฆ **Model Size** | Only **47M** parameters, with an INT8 version of about **50MB** | | ๐Ÿง  **Classification Ability** | The first open-source model to natively support **complete / incomplete / invalid** three-class detection | | ๐Ÿšซ **Invalid Filtering** | Invalid utterance F1 reaches **94.34%**, effectively reducing noise-triggered false activations | | ๐Ÿค— **Open-Source Friendly** | Provides FP32 / INT8 ONNX models, ready to use out of the box |

---
## ๐Ÿ“Œ Introduction **TurnSense** is a **three-class semantic turn detection model** designed for human-machine speech interaction. It focuses on a core problem in conversational systems: > **Should the system respond immediately while the user is speaking, or should it keep waiting?** Traditional approaches usually perform only binary "end-of-turn" detection. **TurnSense goes further** by jointly modeling semantic completeness and invalid input detection. This helps systems achieve more natural turn-taking in complex real-world scenarios and significantly reduces premature interruption, overlapping speech, and invalid triggers.
TurnSense three-class diagram
TurnSense classifies user input into three semantic states: | State | Meaning | Example | | :---: | :--- | :--- | | โœ… **Complete** | The user's expression forms a complete intent, and the system can respond | `"Please check tomorrow's weather in Shanghai."` | | โณ **Incomplete** | The user's expression is not finished and may continue after a pause or truncation | `"I want to ask about that order from yesterday..."` | | ๐Ÿ”‡ **Invalid** | The input does not form valid semantic content and should not trigger a response | `"...(continuous noise / nonverbal vocalization)"` | These three labels allow the system to determine not only **"whether it should take the turn"**, but also **"whether the input is worth responding to"**. This improves interaction naturalness and system stability in voice assistants, real-time calls, intelligent customer service, and other speech interaction scenarios.
---
## โœจ Core Features ### ๐Ÿง  Semantic-Level Three-Class Detection TurnSense jointly models `complete / incomplete / invalid` states. Compared with traditional binary turn detection, this is closer to real conversational behavior. It is also the only open-source solution that natively supports invalid semantic detection. ### โšก Extremely Lightweight and Fast TurnSense has only **47M** parameters. The INT8 version is about **50MB**. In CPU environments, it achieves p50 latency of about **54.65ms** and p90 latency of about **58.00ms**, enabling real-time interaction without requiring a GPU. ### ๐ŸŽฏ Strong Accuracy On `easyturn_real_test_ZH` with 300 samples, TurnSense achieves **F1 96.35%** for `complete` and **F1 96.32%** for `incomplete`. On `semantic_test_ZH` with 2000 samples, it achieves **F1 92.30%** for `complete` and **F1 91.62%** for `incomplete`, reaching best or second-best performance among comparable models. ### ๐Ÿšซ Invalid Input Filtering On the NonverbalVocalization dataset, invalid utterance detection reaches **100% precision**, **90.37% recall**, and **94.34% F1**, effectively suppressing false activations caused by nonverbal vocalizations and noise. ### โš–๏ธ More Robust Turn-Taking Decisions TurnSense balances precision and recall in semantically ambiguous, paused, or colloquial speech scenarios, reducing premature responses and missed responses. ### ๐Ÿ“Š Reproducible Evaluation Pipeline The project includes a complete evaluation workflow and scripts, supporting unified metric comparison and performance regression analysis to ensure reproducibility. ### ๐Ÿค— Open-Source Friendly and Ready to Use TurnSense provides a standardized repository structure and FP32 / INT8 ONNX models. Installation and inference can be completed within minutes.
---
## ๐Ÿ“ Model Size Comparison
| Model | Parameters | Three-Class | Link | | :--- | :---: | :---: | :--- | | TEN-Turn | **7B** | โŒ | [TEN-framework/TEN_Turn_Detection](https://huggingface.co/TEN-framework/TEN_Turn_Detection) | | Easy-Turn | 850M | โŒ | [ASLP-lab/Easy-Turn](https://huggingface.co/ASLP-lab/Easy-Turn) | | NAMO-Turn-Detector (ZH) | 66M | โŒ | [videosdk-live/Namo-Turn-Detector-v1-Multilingual](https://huggingface.co/videosdk-live/Namo-Turn-Detector-v1-Multilingual) | | **โญ TurnSense** | **47M** | **โœ…** | [**Baiji-Team/TurnSense**](https://huggingface.co/brgroup/TurnSense) | | Smart-Turn-v3 | 8M | โŒ | [pipecat-ai/smart-turn-v3](https://huggingface.co/pipecat-ai/smart-turn-v3) | | FireRedChat-turn-detector | -- | โŒ | [FireRedTeam/FireRedChat-turn-detector](https://huggingface.co/FireRedTeam/FireRedChat-turn-detector) |
> ๐Ÿ’ก With only **47M** parameters, TurnSense provides native three-class detection and achieves a strong balance between accuracy and model size.
---
## ๐Ÿ“Š Benchmark Results > The following results cover Chinese, English, and invalid-utterance test sets. Chinese results mainly demonstrate the capability of the initial TurnSense version, while English results show the enhanced performance of TurnSense 1.1.
### ๐Ÿ“‹ easyturn_real_test_ZH๏ผˆ300 samples๏ผ‰ > Data source: real samples from [Easy-Turn-Testset](https://huggingface.co/datasets/ASLP-lab/Easy-Turn-Testset) | Model | P (complete) | R (complete) | **F1 (complete)** | P (incomplete) | R (incomplete) | **F1 (incomplete)** | p50 Latency | p90 Latency | | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | Easy-Turn | 97.26% | 94.67% | 95.95% | 94.81% | 97.33% | 96.05% | 183.87 (GPU) | 300.37 (GPU) | | Smart-Turn-v3 | 64.97% | 76.67% | 70.34% | 71.54% | 58.67% | 64.47% | 36.84 | 39.10 | | TEN-Turn | **99.25%** | 88.00% | 93.29% | 89.22% | **99.33%** | 94.01% | 17.66 (GPU) | 19.41 (GPU) | | FireRedChat | 70.65% | 94.67% | 80.91% | 91.92% | 60.67% | 73.09% | 98.30 | 99.42 | | NAMO-Turn | 81.53% | 85.33% | 83.39% | 84.62% | 80.67% | 82.59% | 3.60 | 83.44 | | **โญ TurnSense** | 96.03% | **96.67%** | **๐Ÿ† 96.35%** | **96.64%** | 96.00% | **๐Ÿ† 96.32%** | 54.65 | 58.00 | > **๐Ÿ” Key finding:** TurnSense achieves the highest F1 for both `complete` and `incomplete`, and is the only model that reaches F1 > 96% with CPU p50 latency below 60ms.
### ๐Ÿ“‹ semantic_test_ZH๏ผˆ2000 samples๏ผ‰ > Data source: Chinese test set from [KE-Team/SemanticVAD-Dataset](https://huggingface.co/datasets/KE-Team/SemanticVAD-Dataset) | Model | P (complete) | R (complete) | **F1 (complete)** | P (incomplete) | R (incomplete) | **F1 (incomplete)** | p50 Latency | p90 Latency | | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | Easy-Turn | 78.14% | 98.30% | 87.07% | 97.64% | 70.30% | 81.74% | 183.87 (GPU) | 300.37 (GPU) | | Smart-Turn-v3 | 59.25% | 88.10% | 70.85% | 76.80% | 39.40% | 52.08% | 36.84 | 39.10 | | TEN-Turn | 85.25% | **99.60%** | 91.87% | **99.52%** | 82.70% | 90.33% | 17.66 (GPU) | 19.41 (GPU) | | FireRedChat | 66.76% | 99.40% | 79.87% | 98.83% | 50.50% | 66.84% | 98.30 | 99.42 | | NAMO-Turn | 71.48% | 86.70% | 78.36% | 83.10% | 65.40% | 73.20% | 3.60 | 83.44 | | **โญ TurnSense** | **88.96%** | 95.90% | **๐Ÿ† 92.30%** | 95.55% | **88.00%** | **๐Ÿ† 91.62%** | 54.65 | 58.00 | > **๐Ÿ” Key finding:** On the larger 2000-sample test set, TurnSense continues to maintain the best F1 performance, demonstrating strong generalization.
### ๐Ÿ“‹ TurnSense 1.1 English Enhancement Results > Model download: [Hugging Face - brgroup/TurnSense](https://huggingface.co/brgroup/TurnSense) > TurnSense 1.1 focuses on improving semantic completeness detection in English scenarios. The following results show its `complete / incomplete` performance on English test sets. #### ten_test_EN | Model | P (complete) | R (complete) | **F1 (complete)** | P (incomplete) | R (incomplete) | **F1 (incomplete)** | | :--- | :---: | :---: | :---: | :---: | :---: | :---: | | Smart-Turn-v3 | 70.66% | 72.46% | 71.55% | 65.05% | 63.02% | 64.02% | | TEN-Turn | **98.61%** | 90.25% | **94.25%** | 89.15% | **98.44%** | **93.56%** | | FireRedChat | 76.41% | **97.46%** | 85.66% | **95.28%** | 63.02% | 75.86% | | NAMO-Turn | 92.65% | 26.69% | 41.45% | 51.94% | 97.40% | 67.75% | | **โญ TurnSense 1.1 int8** | 83.01% | 91.10% | 86.87% | 87.57% | 77.08% | 81.99% | #### semantic_test_EN | Model | P (complete) | R (complete) | **F1 (complete)** | P (incomplete) | R (incomplete) | **F1 (incomplete)** | | :--- | :---: | :---: | :---: | :---: | :---: | :---: | | Smart-Turn-v3 | 68.18% | 75.00% | 71.43% | 72.22% | 65.00% | 68.42% | | TEN-Turn | **97.98%** | 97.00% | **97.49%** | **97.03%** | **98.00%** | **97.51%** | | FireRedChat | 72.06% | **98.00%** | 83.05% | 96.88% | 62.00% | 75.61% | | NAMO-Turn | 93.55% | 87.00% | 90.16% | 87.85% | 94.00% | 90.82% | | **โญ TurnSense 1.1 int8** | 74.60% | 94.00% | 83.19% | 91.89% | 68.00% | 78.16% |
### ๐Ÿ“‹ NonverbalVocalization_invalid๏ผˆ728 samples๏ผ‰ > Data source: OpenSLR [Deeply Nonverbal Vocalization Dataset๏ผˆSLR99๏ผ‰](https://openslr.elda.org/99/) | Model | R (invalid) | | :--- | :---: | | **โญ TurnSense** | **90.37%** | > **๐Ÿ” Key finding:** TurnSense supports invalid semantic detection and can effectively reduce system responses triggered by nonverbal vocalizations or noise.
---
## ๐Ÿš€ Quick Start ### 1. Installation ```bash git clone https://github.com/Bairong-Xdynamics/TurnSense.git cd TurnSense pip install -U numpy onnxruntime torch librosa soundfile pandas scikit-learn huggingface_hub ``` ### 2. Download Model Weights TurnSense model weights are available on Hugging Face: [Baiji-Team/TurnSense](https://huggingface.co/brgroup/TurnSense) | Version | Size | Use Case | | :--- | :--- | :--- | | FP32 | ~191 MB | Accuracy-first scenarios | | INT8 | ~50 MB | Deployment-first scenarios, recommended | **Download options:** **Option 1: Automatic download, recommended** The inference script includes Hugging Face download logic and will automatically download and cache the model during the first run. **Option 2: Git LFS** ```bash git lfs install git clone https://huggingface.co/brgroup/TurnSense ``` **Option 3: Hugging Face Hub** ```python from huggingface_hub import snapshot_download snapshot_download(repo_id="brgroup/TurnSense") ``` ### 3. Inference ```bash python infer.py ``` Example output: ```text Loading model from brgroup/TurnSense... Running inference on: "I want to ask about that order from yesterday..." Results: Input: "I want to ask about that order from yesterday..." TurnSense Detection Result: "incomplete" ```
---
## ๐Ÿงช Evaluation Guide ### 1. Evaluation Pipeline 1. Read test datasets in `.jsonl` format. 2. Warm up each model first. The default value is `warmup_iters=20`. 3. Run inference sample by sample and collect classification and performance metrics. 4. Automatically export summary reports and detailed result files. Output files include: | File | Description | | :--- | :--- | | `report.md` | Summary evaluation report | | `results.json` | Structured evaluation results | | `config.json` | Evaluation configuration | | `per_sample__*.jsonl` | Per-sample prediction results | ### 2. Data Format Requirements๏ผˆJSONL๏ผ‰ Each line should be a JSON object containing at least the following fields: | Field | Description | | :--- | :--- | | `audio_path` | Path to the audio file | | `text` | Text content | | `label` | Label: `complete` / `incomplete` / `invalid` | Example: ```jsonl {"audio_path":"/001.wav","text":"Please check tomorrow's weather in Shanghai.","label":"complete"} {"audio_path":"/002.wav","text":"I want to ask about that order from yesterday...","label":"incomplete"} {"audio_path":"/003.wav","text":"uh... hmm... continuous noise","label":"invalid"} ``` ### 3. Run Evaluation ```bash python TurnSense/Turn_benchmark/benchmark.py ```
---
## ๐Ÿ“š Citation If you use TurnSense in your research or product, please cite: ```bibtex @misc{turnsense2026, author = {Baiji Team}, title = {TurnSense: A Three-Class Semantic Detection Model for Complete, Incomplete, and Invalid Utterances}, year = {2026}, publisher = {Hugging Face}, howpublished = {\url{https://huggingface.co/brgroup/TurnSense}}, } ```

## โ“ Questions and Contact If you have questions or suggestions, feel free to contact us through the following channels: | Channel | Contact | | :--- | :--- | | ๐Ÿ“ง Email | [huan.shen@brgroup.com](mailto:huan.shen@brgroup.com) ใƒป [yingao.wang@brgroup.com](mailto:yingao.wang@brgroup.com) ใƒป [wei.zou@brgroup.com](mailto:wei.zou@brgroup.com) | | ๐Ÿ’ฌ WeChat | h2538406363 | | ๐Ÿ‘ฅ WeChat Group | Scan the QR code to join the group
WeChat group QR code | | ๐Ÿ› Issues | [GitHub Issues](https://github.com/Bairong-Xdynamics/TurnSense/issues) | | ๐Ÿ”€ PR | [Pull Requests](https://github.com/Bairong-Xdynamics/TurnSense/pulls) |
## ๐Ÿ“„ License This project is released under the **Apache License 2.0** with additional specific restrictions. See [LICENSE](./LICENSE) for details.
---
**Built with โค๏ธ by [Baiji Team](https://github.com/Bairong-Xdynamics)**