Chat-based Search for Intellectual Property Data

Sammanfattning

The rapid advancements in artificial intelligence, particularly in natural language processing and large language models, have created new opportunities for improving information retrieval systems. This thesis explores a chat-based search system for intellectual property data. The proposed system leverages large language models to transition from traditional keyword-based searches to conversational interfaces. By integrating retrieval-augmented generation techniques, the system enhances user interaction quality and answer accuracy. The study addresses the architecture design, the implementation, and the evaluation of the chat-based search system, focusing on patents as the primary data source. The final architecture employs a tool-based combined retriever with semantic search and NL2SQL, using GPT-4o as its large language model. User tests show improvements in retrieval efficiency, indicating a promising future for conversational artificial intelligence in intellectual property information retrieval.