| --- |
| license: mit |
| task_categories: |
| - text-retrieval |
| - question-answering |
| - sentence-similarity |
| - text-generation |
| - text-classification |
| language: |
| - vi |
| tags: |
| - sentence-transformers |
| - transformers |
| - PyTorch |
| - law |
| - dense-retrieval |
| - semantic-search |
| - benchmark |
| - domain-specific |
| - mteb |
| size_categories: |
| - 100K<n<1M |
| pretty_name: Vietnamese Legal Documents Dataset |
| --- |
| |
| # Vietnamese Legal Documents Dataset |
|
|
| ## 1. Dataset Summary |
|
|
| **Raw data**: [tmnam20/BKAI-Legal-Retrieval](https://huggingface.co/datasets/tmnam20/BKAI-Legal-Retrieval) |
|
|
| The **Vietnamese Legal Documents Dataset** is a benchmark dataset designed for **legal information retrieval** in the Vietnamese language. It consists of: |
|
|
| - A **corpus** of legal documents. |
| - **Train/test splits** containing natural language queries and their corresponding relevant documents. |
|
|
| This dataset is intended to support research and development in: |
|
|
| - **Information Retrieval (IR)** |
| - **Retrieval-Augmented Generation (RAG)** |
| - **Vietnamese Legal NLP** |
|
|
| --- |
|
|
| ## 2. Supported Tasks and Use Cases |
|
|
| The dataset can be used for the following tasks: |
|
|
| - **Information Retrieval (IR)** |
| Retrieve relevant legal documents given a natural-language query. |
|
|
| - **Question Answering over Documents** |
| Serve as a retrieval component in QA systems, where retrieved documents are used as grounding context for downstream models (e.g., LLMs). |
|
|
| - **Legal NLP Research** |
| Benchmark and evaluate models in a domain-specific, Vietnamese legal setting. |
|
|
| --- |
|
|
| ## 3. Dataset Structure |
|
|
| ### 3.1 Corpus |
|
|
| `corpus.parquet` |
|
|
| | column | type | description | |
| |-------:|:-------|:-------------------------------------------| |
| | `cid` | int | Unique identifier for each document | |
| | `text` | string | Full text of a legal document (Vietnamese) | |
|
|
| ### 3.2 Train / Test Splits |
|
|
| `train.parquet` / `test.parquet` |
|
|
| | column | type | description | |
| |---------------:|:-------------|:----------------------------------------------------------------------------| |
| | `qid` | int | Unique query identifier | |
| | `question` | string | Natural-language legal query in Vietnamese | |
| | `cid` | list<int> | List of relevant document IDs (subset of `corpus.cid`) | |
| | `context_list` | list<string> | List of text passages mapped from `cid` (aligned by index with `cid` field) | |
|
|
| --- |
|
|
| ## 4. **Example Usuage**: |
|
|
| ```python |
| from datasets import load_dataset |
| |
| ds = load_dataset("YuITC/Vietnamese-Legal-Doc-Retrieval-Data") |
| |
| print(ds["train"][0]) |
| # { |
| # "qid": 72600, |
| # "question": "Liên đoàn Luật sư Việt Nam là tổ chức ...", |
| # "context_list": ["Điều 2. Địa vị pháp lý của Liên đoàn Luật sư Việt Nam\n1. Liên đoàn Luật sư ..."] |
| # "cid": [142820] |
| # } |
| ``` |
|
|
| --- |
|
|
| ## 5. Limitations |
|
|
| - The dataset does not fully cover the entire Vietnamese legal system. |
| - Legal texts contain domain-specific terminology, which may be challenging for non-expert users and models. |
