Law pioneers : developing a question answering system in a Chinese legal domain using large language model | |
| Author | Shen, Jiewen |
| Call Number | AIT Thesis no.CS-24-02 |
| Subject(s) | Question-answering systems Natural language generation (Computer science) Legal services |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science |
| Publisher | Asian Institute of Technology |
| Abstract | In recent years, large language models (LLMs) have greatly propelled the development of the field of Natural Language Processing (NLP), especially with the emergence of general-purpose LLMs that have exhibited remarkable achievements across various tra ditional language tasks, showcasing their robust practicality. However, the primary fo cus of general-purpose LLMs lies in their versatility, and addressing complex problems in specific domains, such as legal applications, encounters certain challenges. Direct use of general-purpose large language models may lead to simplistic or overly gener alized answers. Furthermore, specific references in legal applications are something that general language models cannot provide. Therefore, research and development of domain-specific models based on the legal field are essential. Specifically, totacklethese obstacles, there has been a significant increase in research and practices in recent years.Simultaneously recognizing these challenges, we have designed an intelligent legal model named LawPioneer, which is an improvement upon general purpose language models (LLMs) and aims to provide extensive legal services. LawPioneer builds upon general purpose large models with vertical domain optimization and enhancement to improve the accuracy of legal question-answering. This modification aims to achieve significant progress in the field of intelligent legal QA, ensuring that users can more easily under stand and apply the legal information provided by the system. LawPioneer can reference relevant legal statutes when answering questions and mitigates common hallucination issues seen in general language models. This enhancement strives to make significant advancements in the field of legal intelligent question-answering, ensuring users can more easily comprehend and apply the legal information provided by the system.Quantitative and qualitative results of LawPioneer demonstrate that our model achieves effective enhancements in professional accuracy and practical usability over general purpose models. |
| Year | 2024 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Chaklam Silpasuwanchai; |
| Examination Committee(s) | Chantri Polprasert;Attaphongse Taparugssanagorn; |
| Scholarship Donor(s) | China Scholarship Council (CSC); |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2024 |