The paper presents the methodology employed by the TQM team in the COLIEE2024 competition, which focused on legal case retrieval tasks.
The authors first perform fine-grained data pre-processing to eliminate noisy information in the case documents. They then implement classical lexical matching models, such as BM25 and QLD, and state-of-the-art semantic retrieval models, including SAILER and DELTA, with a focus on improving the understanding of case relevance.
To further enhance the modeling of case relevance, the authors utilize learning-to-rank techniques to integrate various features, including scores from the lexical and semantic retrieval models, as well as other metadata-based features.
Additionally, the authors propose heuristic post-processing strategies, such as filtering by trial date, filtering query cases, and filtering duplicate cases, to mitigate the impact of irrelevant information.
The official results of the COLIEE2024 competition reveal the effectiveness of the TQM team's approach, with the team securing first place in Task 1 and third place in Task 3.
The authors anticipate that their proposed methodology can contribute valuable insights to the advancement of legal retrieval technology, particularly in enhancing the understanding of case relevance.
To Another Language
from source content
arxiv.org
Viktige innsikter hentet fra
by Haitao Li,Yo... klokken arxiv.org 04-02-2024
https://arxiv.org/pdf/2404.00947.pdfDypere Spørsmål