Adaptive Two-Phase Finetuning LLMs for Japanese Legal Text Retrieval

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Trung, Quang Hoang, Phuc, Nguyen Van Hoang, Hoang, Le Trung, Hieu, Quang Huu, Duy, Vo Nguyen Le
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912160169328640
author Trung, Quang Hoang
Phuc, Nguyen Van Hoang
Hoang, Le Trung
Hieu, Quang Huu
Duy, Vo Nguyen Le
author_facet Trung, Quang Hoang
Phuc, Nguyen Van Hoang
Hoang, Le Trung
Hieu, Quang Huu
Duy, Vo Nguyen Le
contents Text Retrieval (TR) involves finding and retrieving text-based content relevant to a user's query from a large repository, with applications in real-world scenarios such as legal document retrieval. While most existing studies focus on English, limited work addresses Japanese contexts. In this paper, we introduce a new dataset specifically designed for Japanese legal contexts and propose a novel two-phase pipeline tailored to this domain. In the first phase, the model learns a broad understanding of global contexts, enhancing its generalization and adaptability to diverse queries. In the second phase, the model is fine-tuned to address complex queries specific to legal scenarios. Extensive experiments are conducted to demonstrate the superior performance of our method, which outperforms existing baselines. Furthermore, our pipeline proves effective in English contexts, surpassing comparable baselines on the MS MARCO dataset. We have made our code publicly available on GitHub, and the model checkpoints are accessible via HuggingFace.
format Preprint
id arxiv_https___arxiv_org_abs_2412_13205
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Adaptive Two-Phase Finetuning LLMs for Japanese Legal Text Retrieval
Trung, Quang Hoang
Phuc, Nguyen Van Hoang
Hoang, Le Trung
Hieu, Quang Huu
Duy, Vo Nguyen Le
Information Retrieval
Computation and Language
Machine Learning
Text Retrieval (TR) involves finding and retrieving text-based content relevant to a user's query from a large repository, with applications in real-world scenarios such as legal document retrieval. While most existing studies focus on English, limited work addresses Japanese contexts. In this paper, we introduce a new dataset specifically designed for Japanese legal contexts and propose a novel two-phase pipeline tailored to this domain. In the first phase, the model learns a broad understanding of global contexts, enhancing its generalization and adaptability to diverse queries. In the second phase, the model is fine-tuned to address complex queries specific to legal scenarios. Extensive experiments are conducted to demonstrate the superior performance of our method, which outperforms existing baselines. Furthermore, our pipeline proves effective in English contexts, surpassing comparable baselines on the MS MARCO dataset. We have made our code publicly available on GitHub, and the model checkpoints are accessible via HuggingFace.
title Adaptive Two-Phase Finetuning LLMs for Japanese Legal Text Retrieval
topic Information Retrieval
Computation and Language
Machine Learning
url https://arxiv.org/abs/2412.13205