Improving Vietnamese Legal Document Retrieval using Synthetic Data

Pham Tien Son, Nguyen Doan Hieu, Nguyen Dai An, Dinh Viet Sang · 2024

Read Improving Vietnamese Legal Document Retrieval using Synthetic Data by Pham Tien Son, Nguyen Doan Hieu, Nguyen Dai An, Dinh Viet Sang (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-00657.txt.

Description

Pham Tien Son and three others present a 2024 paper on searching Vietnamese legal texts with AI. Systems that answer questions need search models trained on marked examples, but Vietnamese law lacks large marked data sets. So the team used large language models to make the training set, writing varied, life-like legal queries for real passages.

Two search designs were pre-trained on this made-up data and then fine-tuned with contrastive learning against hard negative cases. Those cases look apt but are not. Search accuracy rose sharply over the base scores. The study shows made-up data can stand in for missing labels in legal settings with scarce data. That is a pattern other niche tongues could copy.

Catalog record

CREATORPham Tien Son, Nguyen Doan Hieu, Nguyen Dai An, Dinh Viet Sang
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-00657

Browse the full library · About the Tower of Babel