<p align="center">
<img alt="SemViQA Logo" src="image/logo.png" height="250" />
<br>
</p>
</p>
# **SemViQA: A Semantic Question Answering System for Vietnamese Information Fact-Checking**
### **Authors**:
[**Nam V. Nguyen**](https://github.com/DAVID-NGUYEN-S16), [**Dien X. Tran**](https://github.com/xndien2004), Thanh T. Tran, Anh T. Hoang, Tai V. Duong, Di T. Le, Phuc-Lu Le
<p align="center">
<a href="https://arxiv.org/abs/2503.00955">
<img src="https://img.shields.io/badge/arXiv-2411.00918-red?style=flat&label=arXiv">
</a>
<a href="https://huggingface.co/SemViQA">
<img src="https://img.shields.io/badge/Hugging%20Face-Model-yellow?style=flat">
</a>
<a href="https://pypi.org/project/SemViQA">
<img src="https://img.shields.io/pypi/v/SemViQA?color=blue&label=PyPI">
</a>
<a href="https://github.com/DAVID-NGUYEN-S16/SemViQA">
<img src="https://img.shields.io/github/stars/DAVID-NGUYEN-S16/SemViQA?style=social">
</a>
</p>
<p align="center">
<a href="#-about">📌 About</a> •
<a href="#-checkpoints">🔍 Checkpoints</a> •
<a href="#-quick-start">🚀 Quick Start</a> •
<a href="#-training">🏋️♂️ Training</a> •
<a href="#-evaluation">📊 Evaluation</a> •
<a href="#-citation">📖 Citation</a>
</p>
---
## 📌 **About**
Misinformation is a growing problem, exacerbated by the increasing use of **Large Language Models (LLMs)** like GPT and Gemini. This issue is even more critical for **low-resource languages like Vietnamese**, where existing fact-checking methods struggle with **semantic ambiguity, homonyms, and complex linguistic structures**.
To address these challenges, we introduce **SemViQA**, a novel **Vietnamese fact-checking framework** integrating:
- **Semantic-based Evidence Retrieval (SER)**: Combines **TF-IDF** with a **Question Answering Token Classifier (QATC)** to enhance retrieval precision while reducing inference time.
- **Two-step Verdict Classification (TVC)**: Uses hierarchical classification optimized with **Cross-Entropy and Focal Loss**, improving claim verification across three categories:
- **Supported** ✅
- **Refuted** ❌
- **Not Enough Information (NEI)** 🤷♂️
### **🏆 Achievements**
- **1st place** in the **UIT Data Science Challenge** 🏅
- **State-of-the-art** performance on:
- **ISE-DSC01** → **78.97% strict accuracy**
- **ViWikiFC** → **80.82% strict accuracy**
- **SemViQA Faster**: **7x speed improvement** over the standard model 🚀
These results establish **SemViQA** as a **benchmark for Vietnamese fact verification**, advancing efforts to combat misinformation and ensure **information integrity**.
---
## 🔍 Checkpoints
We are making our **SemViQA** experiment checkpoints publicly available to support the **Vietnamese fact-checking research community**. By sharing these models, we aim to:
- **Facilitate reproducibility**: Allow researchers and developers to validate and build upon our results.
- **Save computational resources**: Enable fine-tuning or transfer learning on top of **pre-trained and fine-tuned models** instead of training from scratch.
- **Encourage further improvements**: Provide a strong baseline for future advancements in **Vietnamese misinformation detection**.
<table>
<tr>
<th>Method</th>
<th>Model</th>
<th>ViWikiFC</th>
<th>ISE-DSC01</th>
</tr>
<tr>
<td rowspan="3"><strong>TC</strong></td>
<td>InfoXLM<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/tc-infoxlm-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/tc-infoxlm-isedsc01">Link</a></td>
</tr>
<tr>
<td>XLM-R<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/tc-xlmr-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/tc-xlmr-isedsc01">Link</a></td>
</tr>
<tr>
<td>Ernie-M<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/tc-erniem-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/tc-erniem-isedsc01">Link</a></td>
</tr>
<tr>
<td rowspan="3"><strong>BC</strong></td>
<td>InfoXLM<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/bc-infoxlm-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/bc-infoxlm-isedsc01">Link</a></td>
</tr>
<tr>
<td>XLM-R<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/bc-xlmr-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/bc-xlmr-isedsc01">Link</a></td>
</tr>
<tr>
<td>Ernie-M<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/bc-erniem-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/bc-erniem-isedsc01">Link</a></td>
</tr>
<tr>
<td rowspan="2"><strong>QATC</strong></td>
<td>InfoXLM<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/qatc-infoxlm-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/qatc-infoxlm-isedsc01">Link</a></td>
</tr>
<tr>
<td>ViMRC<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/qatc-vimrc-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/qatc-vimrc-isedsc01">Link</a></td>
</tr>
<tr>
<td rowspan="2"><strong>QA origin</strong></td>
<td>InfoXLM<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/infoxlm-large-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/infoxlm-large-isedsc01">Link</a></td>
</tr>
<tr>
<td>ViMRC<sub>large</sub></td>
<td><a href="https://huggingface.co/SemViQA/vi-mrc-large-viwikifc">Link</a></td>
<td><a href="https://huggingface.co/SemViQA/vi-mrc-large-isedsc01">Link</a></td>
</tr>
</table>
---
## 🚀 **Quick Start**
### 📥 **Installation**
#### **1️⃣ Clone this repository**
```bash
git clone https://github.com/DAVID-NGUYEN-S16/SemViQA.git
cd SemViQA
```
#### **2️⃣ Set up Python environment**
We recommend using **Python 3.11** in a virtual environment (`venv`) or **Anaconda**.
**Using `venv`:**
```bash
python -m venv semviqa_env
source semviqa_env/bin/activate # On MacOS/Linux
semviqa_env\Scripts\activate # On Windows
```
**Using `Anaconda`:**
```bash
conda create -n semviqa_env python=3.11 -y
conda activate semviqa_env
```
#### **3️⃣ Install dependencies**
```bash
pip install --upgrade pip
pip install transformers==4.42.3
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu118
pip install -r requirements.txt
```
---
## 🏋️♂️ **Training**
Train different components of **SemViQA** using the provided scripts:
### **1️⃣ Three-Class Classification Training**
Train a three-class claim classification model using the following command:
```bash
bash scripts/tc.sh
```
If you want to fine-tune the model using pre-trained weights, you can initialize it as follows:
```python
# Install semviqa
!pip install semviqa
# Initalize a pipeline
from transformers import AutoTokenizer
from semviqa.tvc.model import ClaimModelForClassification
tokenizer = AutoTokenizer.from_pretrained("SemViQA/tc-infoxlm-viwikifc")
model = ClaimModelForClassification.from_pretrained("SemViQA/tc-infoxlm-viwikifc", num_labels=3)
```
### **2️⃣ Binary Classification Training**
Train a binary classification model using the command below:
```bash
bash scripts/bc.sh
```
To fine-tune the model with existing weights, use the following setup:
```python
# Install semviqa
!pip install semviqa
# Initalize a pipeline
from transformers import AutoTokenizer
from semviqa.tvc.model import ClaimModelForClassification
tokenizer = AutoTokenizer.from_pretrained("SemViQA/bc-infoxlm-viwikifc")
model = ClaimModelForClassification.from_pretrained("SemViQA/bc-infoxlm-viwikifc", num_labels=2)
```
### **3️⃣ QATC Model Training**
Train the Question Answering Token Classifier (QATC) model using the following command:
```bash
bash scripts/qatc.sh
```
To continue training from pre-trained weights, use this setup:
```python
# Install semviqa
!pip install semviqa
# Initalize a pipeline
from transformers import AutoTokenizer
from semviqa.ser.qatc_model import QATCForQuestionAnswering
tokenizer = AutoTokenizer.from_pretrained("SemViQA/qatc-infoxlm-viwikifc")
model = QATCForQuestionAnswering.from_pretrained("SemViQA/qatc-infoxlm-viwikifc")
```
---
## 📊 **Evaluation**
### **1️⃣ Semantic-based Evidence Retrieval**
This module extracts the most relevant evidence from a given context based on a claim. It leverages TF-IDF combined with the QATC model to ensure accurate retrieval.
```python
# Install semviqa package
!pip install semviqa
# Import the ser module
import torch
from transformers import AutoTokenizer
from semviqa.ser.qatc_model import QATCForQuestionAnswering
from semviqa.ser.ser_eval import extract_evidence_tfidf_qatc
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
tokenizer = AutoTokenizer.from_pretrained("SemViQA/qatc-infoxlm-viwikifc")
model = QATCForQuestionAnswering.from_pretrained("SemViQA/qatc-infoxlm-viwikifc")
claim = "Chiến tranh với Campuchia đã kết thúc trước khi Việt Nam thống nhất."
context = "Sau khi thống nhất, Việt Nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh Liên Xô cùng Khối phía Đông, các lệnh cấm vận của Hoa Kỳ, chiến tranh với Campuchia, biên giới giáp Trung Quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng. Năm 1986, Đảng Cộng sản ban hành cải cách đổi mới, tạo điều kiện hình thành kinh tế thị trường và hội nhập sâu rộng. Cải cách đổi mới kết hợp cùng quy mô dân số lớn đưa Việt Nam trở thành một trong những nước đang phát triển có tốc độ tăng trưởng thuộc nhóm nhanh nhất thế giới, được coi là Hổ mới châu Á dù cho vẫn gặp phải những thách thức như tham nhũng, tội phạm gia tăng, ô nhiễm môi trường và phúc lợi xã hội chưa đầy đủ. Ngoài ra, giới bất đồng chính kiến, chính phủ một số nước phương Tây và các tổ chức theo dõi nhân quyền có quan điểm chỉ trích hồ sơ nhân quyền của Việt Nam liên quan đến các vấn đề tôn giáo, kiểm duyệt truyền thông, hạn chế hoạt động ủng hộ nhân quyền cùng các quyền tự do dân sự."
evidence = extract_evidence_tfidf_qatc(
claim, context, model, tokenizer, device, confidence_threshold=0.5, length_ratio_threshold=0.6
)
print(evidence)
# evidence: sau khi thống nhất việt nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh liên xô cùng khối phía đông các lệnh cấm vận của hoa kỳ chiến tranh với campuchia biên giới giáp trung quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng
```
### **2️⃣ Two-step Verdict Classification**
This module performs claim classification using a **two-step approach**:
1. **Three-class classification**: Determines if a claim is **Supported, Refuted, or Not Enough Information (NEI)**.
2. **Binary classification** (if necessary): Further verifies if the claim is **Supported** or **Refuted**.
```python
# Install semviqa package
!pip install semviqa
# Import the tvc module
import torch
from semviqa.tvc.tvc_eval import classify_claim
from semviqa.tvc.model import ClaimModelForClassification
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
tokenizer = AutoTokenizer.from_pretrained("SemViQA/tc-infoxlm-viwikifc")
model_tc = ClaimModelForClassification.from_pretrained("SemViQA/tc-infoxlm-viwikifc", num_labels=3).to(device)
model_bc = ClaimModelForClassification.from_pretrained("SemViQA/bc-infoxlm-viwikifc", num_labels=2).to(device)
claim = "Chiến tranh với Campuchia đã kết thúc trước khi Việt Nam thống nhất."
evidence = "Sau khi thống nhất, Việt Nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh Liên Xô cùng Khối phía Đông, các lệnh cấm vận của Hoa Kỳ, chiến tranh với Campuchia, biên giới giáp Trung Quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng."
verdict = "NEI"
prob_tc, pred_tc = classify_claim(claim, evidence, model_tc, tokenizer, device)
if pred_tc != 0:
prob_bc, pred_bc = classify_claim(claim, evidence, model_bc, tokenizer, device)
verdict = "SUPPORTED" if pred_bc == 0 else "REFUTED" if prob_bc > prob_tc else ["NEI", "SUPPORTED", "REFUTED"][pred_tc]
print(verdict)
# Output: REFUTED
```
### **3️⃣ Full Pipeline Evaluation**
Use the trained models to **predict test data**:
```bash
bash scripts/pipeline.sh
```
Alternatively, you can use **SemViQA** programmatically:
```python
# Install semviqa package
!pip install semviqa
# Import the pipeline
from semviqa.pipeline import SemViQAPipeline
claim = "Chiến tranh với Campuchia đã kết thúc trước khi Việt Nam thống nhất."
context = "Sau khi thống nhất, Việt Nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh Liên Xô cùng Khối phía Đông, các lệnh cấm vận của Hoa Kỳ, chiến tranh với Campuchia, biên giới giáp Trung Quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng. Năm 1986, Đảng Cộng sản ban hành cải cách đổi mới, tạo điều kiện hình thành kinh tế thị trường và hội nhập sâu rộng. Cải cách đổi mới kết hợp cùng quy mô dân số lớn đưa Việt Nam trở thành một trong những nước đang phát triển có tốc độ tăng trưởng thuộc nhóm nhanh nhất thế giới, được coi là Hổ mới châu Á dù cho vẫn gặp phải những thách thức như tham nhũng, tội phạm gia tăng, ô nhiễm môi trường và phúc lợi xã hội chưa đầy đủ. Ngoài ra, giới bất đồng chính kiến, chính phủ một số nước phương Tây và các tổ chức theo dõi nhân quyền có quan điểm chỉ trích hồ sơ nhân quyền của Việt Nam liên quan đến các vấn đề tôn giáo, kiểm duyệt truyền thông, hạn chế hoạt động ủng hộ nhân quyền cùng các quyền tự do dân sự."
semviqa = SemViQAPipeline(
model_evidence_QA="SemViQA/qatc-infoxlm-viwikifc",
model_bc="SemViQA/bc-infoxlm-viwikifc",
model_tc="SemViQA/tc-infoxlm-viwikifc",
thres_evidence=0.5,
length_ratio_threshold=0.5,
is_qatc_faster=False
)
result = semviqa.predict(claim, context)
print(result)
# Output: {'verdict': 'REFUTED', 'evidence': 'sau khi thống nhất việt nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh liên xô cùng khối phía đông các lệnh cấm vận của hoa kỳ chiến tranh với campuchia biên giới giáp trung quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng'}
# Extract only evidence
evidence_only = semviqa.predict(claim, context, return_evidence_only=True)
print(evidence_only)
# Output: {'evidence': 'sau khi thống nhất việt nam tiếp tục gặp khó khăn do sự sụp đổ và tan rã của đồng minh liên xô cùng khối phía đông các lệnh cấm vận của hoa kỳ chiến tranh với campuchia biên giới giáp trung quốc và hậu quả của chính sách bao cấp sau nhiều năm áp dụng'}
```
## **Acknowledgment**
Our development is based on our previous works:
- [Check-Fact-Question-Answering-System](https://github.com/DAVID-NGUYEN-S16/Check-Fact-Question-Answering-System)
- [Extract-Evidence-Question-Answering](https://github.com/DAVID-NGUYEN-S16/Extract-evidence-question-answering)
**SemViQA** is the final version we have developed for verifying fact-checking in Vietnamese, achieving state-of-the-art (SOTA) performance compared to any other system for Vietnamese.
## 📖 **Citation**
If you use **SemViQA** in your research, please cite our work:
```bibtex
@misc{nguyen2025semviqasemanticquestionanswering,
title={SemViQA: A Semantic Question Answering System for Vietnamese Information Fact-Checking},
author={Nam V. Nguyen and Dien X. Tran and Thanh T. Tran and Anh T. Hoang and Tai V. Duong and Di T. Le and Phuc-Lu Le},
year={2025},
eprint={2503.00955},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2503.00955},
}
```
Raw data
{
"_id": null,
"home_page": "https://github.com/DAVID-NGUYEN-S16/SemViQA",
"name": "semviqa",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": "Vietnamese NLP, Fact-Checking, Question Answering, Machine Learning",
"author": "Nam V. Nguyen, Dien X. Tran, Thanh T. Tran, Anh T. Hoang, Tai V. Duong, Di T. Le, Phuc-Lu Le",
"author_email": "xuandienk4@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/13/22/7c4ae8e9620f04a08549f95517c4097697f3dd5beefc5b1194185c2eb92f/semviqa-1.1.1.tar.gz",
"platform": null,
"description": "<p align=\"center\">\r\n <img alt=\"SemViQA Logo\" src=\"image/logo.png\" height=\"250\" /> \r\n <br>\r\n </p>\r\n</p>\r\n\r\n# **SemViQA: A Semantic Question Answering System for Vietnamese Information Fact-Checking** \r\n\r\n### **Authors**: \r\n[**Nam V. Nguyen**](https://github.com/DAVID-NGUYEN-S16), [**Dien X. Tran**](https://github.com/xndien2004), Thanh T. Tran, Anh T. Hoang, Tai V. Duong, Di T. Le, Phuc-Lu Le \r\n<p align=\"center\">\r\n <a href=\"https://arxiv.org/abs/2503.00955\">\r\n <img src=\"https://img.shields.io/badge/arXiv-2411.00918-red?style=flat&label=arXiv\">\r\n </a>\r\n <a href=\"https://huggingface.co/SemViQA\">\r\n <img src=\"https://img.shields.io/badge/Hugging%20Face-Model-yellow?style=flat\">\r\n </a>\r\n <a href=\"https://pypi.org/project/SemViQA\">\r\n <img src=\"https://img.shields.io/pypi/v/SemViQA?color=blue&label=PyPI\">\r\n </a>\r\n <a href=\"https://github.com/DAVID-NGUYEN-S16/SemViQA\">\r\n <img src=\"https://img.shields.io/github/stars/DAVID-NGUYEN-S16/SemViQA?style=social\">\r\n </a>\r\n</p>\r\n\r\n\r\n<p align=\"center\">\r\n <a href=\"#-about\">\ud83d\udccc About</a> \u2022\r\n <a href=\"#-checkpoints\">\ud83d\udd0d Checkpoints</a> \u2022\r\n <a href=\"#-quick-start\">\ud83d\ude80 Quick Start</a> \u2022\r\n <a href=\"#-training\">\ud83c\udfcb\ufe0f\u200d\u2642\ufe0f Training</a> \u2022\r\n <a href=\"#-evaluation\">\ud83d\udcca Evaluation</a> \u2022\r\n <a href=\"#-citation\">\ud83d\udcd6 Citation</a>\r\n</p> \r\n\r\n---\r\n\r\n## \ud83d\udccc **About** \r\n\r\nMisinformation is a growing problem, exacerbated by the increasing use of **Large Language Models (LLMs)** like GPT and Gemini. This issue is even more critical for **low-resource languages like Vietnamese**, where existing fact-checking methods struggle with **semantic ambiguity, homonyms, and complex linguistic structures**. \r\n\r\nTo address these challenges, we introduce **SemViQA**, a novel **Vietnamese fact-checking framework** integrating: \r\n\r\n- **Semantic-based Evidence Retrieval (SER)**: Combines **TF-IDF** with a **Question Answering Token Classifier (QATC)** to enhance retrieval precision while reducing inference time. \r\n- **Two-step Verdict Classification (TVC)**: Uses hierarchical classification optimized with **Cross-Entropy and Focal Loss**, improving claim verification across three categories: \r\n - **Supported** \u2705 \r\n - **Refuted** \u274c \r\n - **Not Enough Information (NEI)** \ud83e\udd37\u200d\u2642\ufe0f \r\n\r\n### **\ud83c\udfc6 Achievements**\r\n- **1st place** in the **UIT Data Science Challenge** \ud83c\udfc5 \r\n- **State-of-the-art** performance on: \r\n - **ISE-DSC01** \u2192 **78.97% strict accuracy** \r\n - **ViWikiFC** \u2192 **80.82% strict accuracy** \r\n- **SemViQA Faster**: **7x speed improvement** over the standard model \ud83d\ude80 \r\n\r\nThese results establish **SemViQA** as a **benchmark for Vietnamese fact verification**, advancing efforts to combat misinformation and ensure **information integrity**. \r\n\r\n---\r\n## \ud83d\udd0d Checkpoints\r\nWe are making our **SemViQA** experiment checkpoints publicly available to support the **Vietnamese fact-checking research community**. By sharing these models, we aim to: \r\n\r\n- **Facilitate reproducibility**: Allow researchers and developers to validate and build upon our results. \r\n- **Save computational resources**: Enable fine-tuning or transfer learning on top of **pre-trained and fine-tuned models** instead of training from scratch. \r\n- **Encourage further improvements**: Provide a strong baseline for future advancements in **Vietnamese misinformation detection**. \r\n \r\n\r\n<table>\r\n <tr>\r\n <th>Method</th>\r\n <th>Model</th>\r\n <th>ViWikiFC</th>\r\n <th>ISE-DSC01</th>\r\n </tr>\r\n <tr>\r\n <td rowspan=\"3\"><strong>TC</strong></td>\r\n <td>InfoXLM<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-infoxlm-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-infoxlm-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>XLM-R<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-xlmr-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-xlmr-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>Ernie-M<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-erniem-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/tc-erniem-isedsc01\">Link</a></td> \r\n </tr>\r\n <tr>\r\n <td rowspan=\"3\"><strong>BC</strong></td>\r\n <td>InfoXLM<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-infoxlm-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-infoxlm-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>XLM-R<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-xlmr-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-xlmr-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>Ernie-M<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-erniem-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/bc-erniem-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td rowspan=\"2\"><strong>QATC</strong></td>\r\n <td>InfoXLM<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/qatc-infoxlm-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/qatc-infoxlm-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>ViMRC<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/qatc-vimrc-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/qatc-vimrc-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td rowspan=\"2\"><strong>QA origin</strong></td>\r\n <td>InfoXLM<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/infoxlm-large-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/infoxlm-large-isedsc01\">Link</a></td>\r\n </tr>\r\n <tr>\r\n <td>ViMRC<sub>large</sub></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/vi-mrc-large-viwikifc\">Link</a></td>\r\n <td><a href=\"https://huggingface.co/SemViQA/vi-mrc-large-isedsc01\">Link</a></td>\r\n </tr>\r\n</table>\r\n\r\n \r\n\r\n---\r\n\r\n## \ud83d\ude80 **Quick Start** \r\n\r\n### \ud83d\udce5 **Installation** \r\n\r\n#### **1\ufe0f\u20e3 Clone this repository** \r\n```bash\r\ngit clone https://github.com/DAVID-NGUYEN-S16/SemViQA.git\r\ncd SemViQA\r\n```\r\n\r\n#### **2\ufe0f\u20e3 Set up Python environment** \r\nWe recommend using **Python 3.11** in a virtual environment (`venv`) or **Anaconda**. \r\n\r\n**Using `venv`:** \r\n```bash\r\npython -m venv semviqa_env\r\nsource semviqa_env/bin/activate # On MacOS/Linux\r\nsemviqa_env\\Scripts\\activate # On Windows\r\n```\r\n\r\n**Using `Anaconda`:** \r\n```bash\r\nconda create -n semviqa_env python=3.11 -y\r\nconda activate semviqa_env\r\n```\r\n\r\n#### **3\ufe0f\u20e3 Install dependencies** \r\n```bash\r\npip install --upgrade pip\r\npip install transformers==4.42.3\r\npip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu118\r\npip install -r requirements.txt\r\n```\r\n---\r\n\r\n## \ud83c\udfcb\ufe0f\u200d\u2642\ufe0f **Training** \r\n\r\nTrain different components of **SemViQA** using the provided scripts: \r\n\r\n### **1\ufe0f\u20e3 Three-Class Classification Training** \r\nTrain a three-class claim classification model using the following command:\r\n```bash\r\nbash scripts/tc.sh\r\n```\r\nIf you want to fine-tune the model using pre-trained weights, you can initialize it as follows:\r\n```python\r\n# Install semviqa\r\n!pip install semviqa\r\n\r\n# Initalize a pipeline\r\nfrom transformers import AutoTokenizer\r\nfrom semviqa.tvc.model import ClaimModelForClassification\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"SemViQA/tc-infoxlm-viwikifc\")\r\nmodel = ClaimModelForClassification.from_pretrained(\"SemViQA/tc-infoxlm-viwikifc\", num_labels=3)\r\n```\r\n\r\n### **2\ufe0f\u20e3 Binary Classification Training** \r\nTrain a binary classification model using the command below:\r\n```bash\r\nbash scripts/bc.sh\r\n```\r\nTo fine-tune the model with existing weights, use the following setup:\r\n```python\r\n# Install semviqa\r\n!pip install semviqa\r\n\r\n# Initalize a pipeline\r\nfrom transformers import AutoTokenizer\r\nfrom semviqa.tvc.model import ClaimModelForClassification\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"SemViQA/bc-infoxlm-viwikifc\")\r\nmodel = ClaimModelForClassification.from_pretrained(\"SemViQA/bc-infoxlm-viwikifc\", num_labels=2)\r\n```\r\n\r\n### **3\ufe0f\u20e3 QATC Model Training** \r\nTrain the Question Answering Token Classifier (QATC) model using the following command:\r\n```bash\r\nbash scripts/qatc.sh\r\n```\r\n\r\nTo continue training from pre-trained weights, use this setup:\r\n```python\r\n# Install semviqa\r\n!pip install semviqa\r\n\r\n# Initalize a pipeline\r\nfrom transformers import AutoTokenizer\r\nfrom semviqa.ser.qatc_model import QATCForQuestionAnswering\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"SemViQA/qatc-infoxlm-viwikifc\")\r\nmodel = QATCForQuestionAnswering.from_pretrained(\"SemViQA/qatc-infoxlm-viwikifc\")\r\n```\r\n\r\n---\r\n\r\n## \ud83d\udcca **Evaluation**\r\n\r\n### **1\ufe0f\u20e3 Semantic-based Evidence Retrieval**\r\nThis module extracts the most relevant evidence from a given context based on a claim. It leverages TF-IDF combined with the QATC model to ensure accurate retrieval.\r\n```python\r\n# Install semviqa package\r\n!pip install semviqa\r\n\r\n# Import the ser module\r\nimport torch\r\nfrom transformers import AutoTokenizer\r\nfrom semviqa.ser.qatc_model import QATCForQuestionAnswering\r\nfrom semviqa.ser.ser_eval import extract_evidence_tfidf_qatc\r\n\r\ndevice = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"SemViQA/qatc-infoxlm-viwikifc\")\r\nmodel = QATCForQuestionAnswering.from_pretrained(\"SemViQA/qatc-infoxlm-viwikifc\")\r\n\r\nclaim = \"Chi\u1ebfn tranh v\u1edbi Campuchia \u0111\u00e3 k\u1ebft th\u00fac tr\u01b0\u1edbc khi Vi\u1ec7t Nam th\u1ed1ng nh\u1ea5t.\"\r\ncontext = \"Sau khi th\u1ed1ng nh\u1ea5t, Vi\u1ec7t Nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh Li\u00ean X\u00f4 c\u00f9ng Kh\u1ed1i ph\u00eda \u0110\u00f4ng, c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a Hoa K\u1ef3, chi\u1ebfn tranh v\u1edbi Campuchia, bi\u00ean gi\u1edbi gi\u00e1p Trung Qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng. N\u0103m 1986, \u0110\u1ea3ng C\u1ed9ng s\u1ea3n ban h\u00e0nh c\u1ea3i c\u00e1ch \u0111\u1ed5i m\u1edbi, t\u1ea1o \u0111i\u1ec1u ki\u1ec7n h\u00ecnh th\u00e0nh kinh t\u1ebf th\u1ecb tr\u01b0\u1eddng v\u00e0 h\u1ed9i nh\u1eadp s\u00e2u r\u1ed9ng. C\u1ea3i c\u00e1ch \u0111\u1ed5i m\u1edbi k\u1ebft h\u1ee3p c\u00f9ng quy m\u00f4 d\u00e2n s\u1ed1 l\u1edbn \u0111\u01b0a Vi\u1ec7t Nam tr\u1edf th\u00e0nh m\u1ed9t trong nh\u1eefng n\u01b0\u1edbc \u0111ang ph\u00e1t tri\u1ec3n c\u00f3 t\u1ed1c \u0111\u1ed9 t\u0103ng tr\u01b0\u1edfng thu\u1ed9c nh\u00f3m nhanh nh\u1ea5t th\u1ebf gi\u1edbi, \u0111\u01b0\u1ee3c coi l\u00e0 H\u1ed5 m\u1edbi ch\u00e2u \u00c1 d\u00f9 cho v\u1eabn g\u1eb7p ph\u1ea3i nh\u1eefng th\u00e1ch th\u1ee9c nh\u01b0 tham nh\u0169ng, t\u1ed9i ph\u1ea1m gia t\u0103ng, \u00f4 nhi\u1ec5m m\u00f4i tr\u01b0\u1eddng v\u00e0 ph\u00fac l\u1ee3i x\u00e3 h\u1ed9i ch\u01b0a \u0111\u1ea7y \u0111\u1ee7. Ngo\u00e0i ra, gi\u1edbi b\u1ea5t \u0111\u1ed3ng ch\u00ednh ki\u1ebfn, ch\u00ednh ph\u1ee7 m\u1ed9t s\u1ed1 n\u01b0\u1edbc ph\u01b0\u01a1ng T\u00e2y v\u00e0 c\u00e1c t\u1ed5 ch\u1ee9c theo d\u00f5i nh\u00e2n quy\u1ec1n c\u00f3 quan \u0111i\u1ec3m ch\u1ec9 tr\u00edch h\u1ed3 s\u01a1 nh\u00e2n quy\u1ec1n c\u1ee7a Vi\u1ec7t Nam li\u00ean quan \u0111\u1ebfn c\u00e1c v\u1ea5n \u0111\u1ec1 t\u00f4n gi\u00e1o, ki\u1ec3m duy\u1ec7t truy\u1ec1n th\u00f4ng, h\u1ea1n ch\u1ebf ho\u1ea1t \u0111\u1ed9ng \u1ee7ng h\u1ed9 nh\u00e2n quy\u1ec1n c\u00f9ng c\u00e1c quy\u1ec1n t\u1ef1 do d\u00e2n s\u1ef1.\"\r\n\r\nevidence = extract_evidence_tfidf_qatc(\r\n claim, context, model, tokenizer, device, confidence_threshold=0.5, length_ratio_threshold=0.6\r\n)\r\n\r\nprint(evidence)\r\n# evidence: sau khi th\u1ed1ng nh\u1ea5t vi\u1ec7t nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh li\u00ean x\u00f4 c\u00f9ng kh\u1ed1i ph\u00eda \u0111\u00f4ng c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a hoa k\u1ef3 chi\u1ebfn tranh v\u1edbi campuchia bi\u00ean gi\u1edbi gi\u00e1p trung qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng\r\n```\r\n\r\n\r\n### **2\ufe0f\u20e3 Two-step Verdict Classification**\r\nThis module performs claim classification using a **two-step approach**:\r\n1. **Three-class classification**: Determines if a claim is **Supported, Refuted, or Not Enough Information (NEI)**.\r\n2. **Binary classification** (if necessary): Further verifies if the claim is **Supported** or **Refuted**.\r\n```python\r\n# Install semviqa package\r\n!pip install semviqa\r\n\r\n# Import the tvc module\r\nimport torch\r\nfrom semviqa.tvc.tvc_eval import classify_claim\r\nfrom semviqa.tvc.model import ClaimModelForClassification\r\n\r\ndevice = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\r\n\r\ntokenizer = AutoTokenizer.from_pretrained(\"SemViQA/tc-infoxlm-viwikifc\")\r\nmodel_tc = ClaimModelForClassification.from_pretrained(\"SemViQA/tc-infoxlm-viwikifc\", num_labels=3).to(device)\r\nmodel_bc = ClaimModelForClassification.from_pretrained(\"SemViQA/bc-infoxlm-viwikifc\", num_labels=2).to(device)\r\n\r\nclaim = \"Chi\u1ebfn tranh v\u1edbi Campuchia \u0111\u00e3 k\u1ebft th\u00fac tr\u01b0\u1edbc khi Vi\u1ec7t Nam th\u1ed1ng nh\u1ea5t.\"\r\nevidence = \"Sau khi th\u1ed1ng nh\u1ea5t, Vi\u1ec7t Nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh Li\u00ean X\u00f4 c\u00f9ng Kh\u1ed1i ph\u00eda \u0110\u00f4ng, c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a Hoa K\u1ef3, chi\u1ebfn tranh v\u1edbi Campuchia, bi\u00ean gi\u1edbi gi\u00e1p Trung Qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng.\"\r\n\r\nverdict = \"NEI\"\r\nprob_tc, pred_tc = classify_claim(claim, evidence, model_tc, tokenizer, device)\r\n\r\nif pred_tc != 0:\r\n prob_bc, pred_bc = classify_claim(claim, evidence, model_bc, tokenizer, device)\r\n verdict = \"SUPPORTED\" if pred_bc == 0 else \"REFUTED\" if prob_bc > prob_tc else [\"NEI\", \"SUPPORTED\", \"REFUTED\"][pred_tc]\r\n\r\nprint(verdict)\r\n# Output: REFUTED\r\n```\r\n\r\n### **3\ufe0f\u20e3 Full Pipeline Evaluation**\r\nUse the trained models to **predict test data**: \r\n```bash\r\nbash scripts/pipeline.sh\r\n```\r\n\r\nAlternatively, you can use **SemViQA** programmatically:\r\n\r\n```python\r\n# Install semviqa package\r\n!pip install semviqa\r\n\r\n# Import the pipeline\r\nfrom semviqa.pipeline import SemViQAPipeline\r\nclaim = \"Chi\u1ebfn tranh v\u1edbi Campuchia \u0111\u00e3 k\u1ebft th\u00fac tr\u01b0\u1edbc khi Vi\u1ec7t Nam th\u1ed1ng nh\u1ea5t.\"\r\ncontext = \"Sau khi th\u1ed1ng nh\u1ea5t, Vi\u1ec7t Nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh Li\u00ean X\u00f4 c\u00f9ng Kh\u1ed1i ph\u00eda \u0110\u00f4ng, c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a Hoa K\u1ef3, chi\u1ebfn tranh v\u1edbi Campuchia, bi\u00ean gi\u1edbi gi\u00e1p Trung Qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng. N\u0103m 1986, \u0110\u1ea3ng C\u1ed9ng s\u1ea3n ban h\u00e0nh c\u1ea3i c\u00e1ch \u0111\u1ed5i m\u1edbi, t\u1ea1o \u0111i\u1ec1u ki\u1ec7n h\u00ecnh th\u00e0nh kinh t\u1ebf th\u1ecb tr\u01b0\u1eddng v\u00e0 h\u1ed9i nh\u1eadp s\u00e2u r\u1ed9ng. C\u1ea3i c\u00e1ch \u0111\u1ed5i m\u1edbi k\u1ebft h\u1ee3p c\u00f9ng quy m\u00f4 d\u00e2n s\u1ed1 l\u1edbn \u0111\u01b0a Vi\u1ec7t Nam tr\u1edf th\u00e0nh m\u1ed9t trong nh\u1eefng n\u01b0\u1edbc \u0111ang ph\u00e1t tri\u1ec3n c\u00f3 t\u1ed1c \u0111\u1ed9 t\u0103ng tr\u01b0\u1edfng thu\u1ed9c nh\u00f3m nhanh nh\u1ea5t th\u1ebf gi\u1edbi, \u0111\u01b0\u1ee3c coi l\u00e0 H\u1ed5 m\u1edbi ch\u00e2u \u00c1 d\u00f9 cho v\u1eabn g\u1eb7p ph\u1ea3i nh\u1eefng th\u00e1ch th\u1ee9c nh\u01b0 tham nh\u0169ng, t\u1ed9i ph\u1ea1m gia t\u0103ng, \u00f4 nhi\u1ec5m m\u00f4i tr\u01b0\u1eddng v\u00e0 ph\u00fac l\u1ee3i x\u00e3 h\u1ed9i ch\u01b0a \u0111\u1ea7y \u0111\u1ee7. Ngo\u00e0i ra, gi\u1edbi b\u1ea5t \u0111\u1ed3ng ch\u00ednh ki\u1ebfn, ch\u00ednh ph\u1ee7 m\u1ed9t s\u1ed1 n\u01b0\u1edbc ph\u01b0\u01a1ng T\u00e2y v\u00e0 c\u00e1c t\u1ed5 ch\u1ee9c theo d\u00f5i nh\u00e2n quy\u1ec1n c\u00f3 quan \u0111i\u1ec3m ch\u1ec9 tr\u00edch h\u1ed3 s\u01a1 nh\u00e2n quy\u1ec1n c\u1ee7a Vi\u1ec7t Nam li\u00ean quan \u0111\u1ebfn c\u00e1c v\u1ea5n \u0111\u1ec1 t\u00f4n gi\u00e1o, ki\u1ec3m duy\u1ec7t truy\u1ec1n th\u00f4ng, h\u1ea1n ch\u1ebf ho\u1ea1t \u0111\u1ed9ng \u1ee7ng h\u1ed9 nh\u00e2n quy\u1ec1n c\u00f9ng c\u00e1c quy\u1ec1n t\u1ef1 do d\u00e2n s\u1ef1.\"\r\n \r\nsemviqa = SemViQAPipeline(\r\n model_evidence_QA=\"SemViQA/qatc-infoxlm-viwikifc\", \r\n model_bc=\"SemViQA/bc-infoxlm-viwikifc\", \r\n model_tc=\"SemViQA/tc-infoxlm-viwikifc\", \r\n thres_evidence=0.5,\r\n length_ratio_threshold=0.5,\r\n is_qatc_faster=False\r\n )\r\n \r\nresult = semviqa.predict(claim, context)\r\nprint(result)\r\n# Output: {'verdict': 'REFUTED', 'evidence': 'sau khi th\u1ed1ng nh\u1ea5t vi\u1ec7t nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh li\u00ean x\u00f4 c\u00f9ng kh\u1ed1i ph\u00eda \u0111\u00f4ng c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a hoa k\u1ef3 chi\u1ebfn tranh v\u1edbi campuchia bi\u00ean gi\u1edbi gi\u00e1p trung qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng'}\r\n\r\n# Extract only evidence\r\nevidence_only = semviqa.predict(claim, context, return_evidence_only=True)\r\nprint(evidence_only)\r\n# Output: {'evidence': 'sau khi th\u1ed1ng nh\u1ea5t vi\u1ec7t nam ti\u1ebfp t\u1ee5c g\u1eb7p kh\u00f3 kh\u0103n do s\u1ef1 s\u1ee5p \u0111\u1ed5 v\u00e0 tan r\u00e3 c\u1ee7a \u0111\u1ed3ng minh li\u00ean x\u00f4 c\u00f9ng kh\u1ed1i ph\u00eda \u0111\u00f4ng c\u00e1c l\u1ec7nh c\u1ea5m v\u1eadn c\u1ee7a hoa k\u1ef3 chi\u1ebfn tranh v\u1edbi campuchia bi\u00ean gi\u1edbi gi\u00e1p trung qu\u1ed1c v\u00e0 h\u1eadu qu\u1ea3 c\u1ee7a ch\u00ednh s\u00e1ch bao c\u1ea5p sau nhi\u1ec1u n\u0103m \u00e1p d\u1ee5ng'}\r\n```\r\n\r\n## **Acknowledgment** \r\nOur development is based on our previous works: \r\n- [Check-Fact-Question-Answering-System](https://github.com/DAVID-NGUYEN-S16/Check-Fact-Question-Answering-System) \r\n- [Extract-Evidence-Question-Answering](https://github.com/DAVID-NGUYEN-S16/Extract-evidence-question-answering) \r\n\r\n**SemViQA** is the final version we have developed for verifying fact-checking in Vietnamese, achieving state-of-the-art (SOTA) performance compared to any other system for Vietnamese.\r\n\r\n## \ud83d\udcd6 **Citation** \r\n\r\nIf you use **SemViQA** in your research, please cite our work: \r\n\r\n```bibtex\r\n@misc{nguyen2025semviqasemanticquestionanswering,\r\n title={SemViQA: A Semantic Question Answering System for Vietnamese Information Fact-Checking}, \r\n author={Nam V. Nguyen and Dien X. Tran and Thanh T. Tran and Anh T. Hoang and Tai V. Duong and Di T. Le and Phuc-Lu Le},\r\n year={2025},\r\n eprint={2503.00955},\r\n archivePrefix={arXiv},\r\n primaryClass={cs.CL},\r\n url={https://arxiv.org/abs/2503.00955}, \r\n}\r\n``` \r\n",
"bugtrack_url": null,
"license": null,
"summary": "SemViQA: A Semantic Question Answering System for Vietnamese Fact-Checking",
"version": "1.1.1",
"project_urls": {
"Competition Results": "https://codalab.lisn.upsaclay.fr/competitions/15497#results",
"Homepage": "https://github.com/DAVID-NGUYEN-S16/SemViQA",
"Paper": "https://arxiv.org/abs/2503.00955",
"Repository": "https://github.com/DAVID-NGUYEN-S16/SemViQA"
},
"split_keywords": [
"vietnamese nlp",
" fact-checking",
" question answering",
" machine learning"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "bcba5d837da014b48d0ee6225aa4a6e1c764e621cba1447593bb11eea8ad4382",
"md5": "0b634e9110864d6af6aa76372b11192d",
"sha256": "954144536173b38f68d8b26176979c207f9889374474dad6416e1fa61c491718"
},
"downloads": -1,
"filename": "semviqa-1.1.1-py2.py3-none-any.whl",
"has_sig": false,
"md5_digest": "0b634e9110864d6af6aa76372b11192d",
"packagetype": "bdist_wheel",
"python_version": "py2.py3",
"requires_python": ">=3.8",
"size": 26427,
"upload_time": "2025-03-09T12:59:47",
"upload_time_iso_8601": "2025-03-09T12:59:47.600804Z",
"url": "https://files.pythonhosted.org/packages/bc/ba/5d837da014b48d0ee6225aa4a6e1c764e621cba1447593bb11eea8ad4382/semviqa-1.1.1-py2.py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "13227c4ae8e9620f04a08549f95517c4097697f3dd5beefc5b1194185c2eb92f",
"md5": "7dc9224febb9bc48e3f56122f34456ab",
"sha256": "1ad7f2a1e9dbe60362ccc9c2f618b0e164fd2c5200ef0e896af0cfc83611c1d2"
},
"downloads": -1,
"filename": "semviqa-1.1.1.tar.gz",
"has_sig": false,
"md5_digest": "7dc9224febb9bc48e3f56122f34456ab",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 28112,
"upload_time": "2025-03-09T12:59:49",
"upload_time_iso_8601": "2025-03-09T12:59:49.197371Z",
"url": "https://files.pythonhosted.org/packages/13/22/7c4ae8e9620f04a08549f95517c4097697f3dd5beefc5b1194185c2eb92f/semviqa-1.1.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-03-09 12:59:49",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "DAVID-NGUYEN-S16",
"github_project": "SemViQA",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [
{
"name": "transformers",
"specs": [
[
"==",
"4.42.3"
]
]
},
{
"name": "datasets",
"specs": [
[
"==",
"2.20.0"
]
]
},
{
"name": "pandas",
"specs": [
[
"==",
"2.2.2"
]
]
},
{
"name": "numpy",
"specs": [
[
"==",
"1.26.4"
]
]
},
{
"name": "underthesea",
"specs": [
[
"==",
"6.8.4"
]
]
},
{
"name": "gc-python-utils",
"specs": [
[
"==",
"0.0.1"
]
]
},
{
"name": "tqdm",
"specs": [
[
"==",
"4.66.4"
]
]
},
{
"name": "safetensors",
"specs": [
[
"==",
"0.4.3"
]
]
},
{
"name": "sentence-transformers",
"specs": [
[
"==",
"3.0.1"
]
]
},
{
"name": "scikit-learn",
"specs": [
[
"==",
"1.2.2"
]
]
},
{
"name": "matplotlib",
"specs": [
[
"==",
"3.7.5"
]
]
},
{
"name": "accelerate",
"specs": [
[
"==",
"0.32.1"
]
]
},
{
"name": "omegaconf",
"specs": [
[
"==",
"2.3.0"
]
]
},
{
"name": "einops",
"specs": [
[
"==",
"0.8.0"
]
]
},
{
"name": "rank_bm25",
"specs": [
[
"==",
"0.2.2"
]
]
},
{
"name": "sentencepiece",
"specs": [
[
"==",
"0.2.0"
]
]
},
{
"name": "pyvi",
"specs": [
[
"==",
"0.1.1"
]
]
}
],
"lcname": "semviqa"
}
JSON
