An Intelligent Price Validation System Based on Large Language Models (LLMs)

Abstract:

This paper presents an innovative approach to car spare parts pricing using an Intelligent Price Validation System based on Large Language Models (LLMs). Traditional pricing methods using conventional strategies for car parts often face challenges such as being very labour intensive with manual errors, fragmented market data, and inefficiencies in processing large datasets. To surmount all these, the proposed system employs a comprehensive approach that merges web scraping and crawling techniques to fetch d ata from the web in real time, based on a Retrieval Augmented Generation
(RAG) schema for rendering accurate responses and the GPT 4 model for advanced natural language processing and natural content generation. The system answers user requests by matching snippets of keywords and phrases in the user request with previously catalogued information on automobile spare parts and then combines, standardizes, and analyses prices using a data processing system developed in Python. The system adopts the use of Fas tAPI on the back end to handle asynchronous processing of multiple simultaneous requests as fast as possible. The search and query module of the system directly matches the user inputs to the predefined keywords using basic string matching techniques, whil e the
data acquisition activities of the system focus on real time aggregation and processing of information from various internet sources. The system was tested using the UTAUT evaluation. The results of the evaluation showed that the system can execute t he requested communications in a few seconds; the system improves users ’ efficiency in finding and validating car part prices, has high accuracy and consistency in performing the task, and is straightforward to use, thereby improving the severity of the se arch and purchasing processes for clients.

Keywords: Large Language Models (LLM), Car spare parts, FastAPI, Market trends, Price validation, Real time automation, Retrieval Augmented Generation, web scraping