README.md

metadata

license: mit
task_categories:
  - text-generation
language:
  - vi
tags:
  - legal
size_categories:
  - 1B<n<10B

TVPL (thuvienphapluat.vn)

2. structured_data_doc.parquet: preprocessed version from only needed doc from tvpl . Please read by Datasets library
3. parent_nodes.parquet: parent nodes from [1] by chunking with SentenceSplitter, chunk_overlap=0, chunk_size=800, tokenizer="Viet-Mistral/Vistral-7B-Chat"
4. child_nodes.parquet: child nodes from [2] by chunking with SentenceSplitter, chunk_overlap=30, chunk_size=190 and using Word Segmentation, vietnamese-bi-encoder

Dedup

WARNING: IF YOU TRAIN PLEASE ONLY CONSIDER dedup/filtered_corpus.parquet and dedup/newtraintestdivide/filtered_corpus.parquet

1. filtered_corpus.parquet: Merged dataset SFT-Law+TVPL-structured+Zalo-corpus. Clustered and then filtered. Items with the longest text are kept.
   • text: str
   • oid: Item's unique ID among 3 merged dataset, int
   • __cluster__: int
   • dataset: Original dataset where the item is extracted from, str
   • __cluster_member__: oid of cluster members from the original 3.

eg:

{
  "text": "Thông tư này hướng dẫn tuần...",
  "dataset": "zalo_legal_corpus",
  "oid": 0,
  "__cluster__": 0
}

2. newtraintestdivide folder contains newly divided SFT_law with 10k tests. oid and __cluster__ has the same meaning as in filtered_corpus.parquet.

3. data_remapped/{file_name}.parquet: data files taken from other repoes, added 2 fields: oid (Item's unique ID among 3 merged dataset, int) and __cluster__ (Cluster ID, int).

   • sft_test+sft_train taken from SFT-Law

   {
   "reference": [
       "https://thuvienphapluat.vn/..."
   ],
   "answer": "Sinh viên học nghệ thuật ca trù ....",
   "question": "Sinh viên học nghệ thuật ca trù tại ...?",
   "domain": ["Tài chính nhà nước"],
   "text": "Căn cứ tại khoản 1 Điều 3 T....",
   "oid": 61425,
   "__cluster__": 219110
   }
   

   • tvpl_dataset taken from TVPL-structured (only the structured_data_doc.parquet file)

   {
   "meta_data": {
       "base": "Căn cứ Luật Tổ chức Chính phủ ngày 19 tháng 6 năm ...",
       "content": "QUY ĐỊNH XỬ PHẠT VI PHẠM HÀNH CHÍNH...",
       "date": "31/12/2021",
       "department": "Chính phủ",
       "doc_type": "Nghị định",
       "file_name": "Decree_No._139_2021_ND-CP_dated_December_31,_2021_.json",
       "id_doc": "139/2021/ND-CP",
       "location": "Hà Nội",
       "title": "Decree No. 139/2021/ND-CP dated December 31, 2021 on Administrative penalties for inland waterway navigation violations",
       "updated": 1710417820
   },
   "child_data": [
       {
       "__cluster__": 253403,
       "header": [
           "Decree No. 139/2021/ND-CP dated...",
           "Chương I. NHỮNG QUY ĐỊNH CHUNG"
       ],
       "len_tokenizer": 157,
       "lower_segmented_text": "điều 1 . phạm_vi điều_chỉnh...",
       "oid": 232323,
       "pointer_link": ["Chương I", "Điều 1"],
       "text": "Điều 1. Phạm vi điều chỉnh\n1. ..."
       }
       ]
   }
   

   • zalo_legal_corpus from Zalo-corpus (only the legal_corpus.json file). law_id is append to articles.

   {
   "article_id": "1",
   "text": "Thông tư này hướng dẫn tuần tra,...",
   "title": "Điều 1. Phạm vi áp dụng",
   "law_id": "01/2009/tt-bnn",
   "oid": 0,
   "__cluster__": 0
   }
   

Please refer to the corresponding repositories for document on the data.