Research on Automatic Question Answering System

Abstract

: With the rapid development of artificial intelligence technology, automatic question answering system is becoming more and more important in the field of natural language processing. However, the existing systems are confronted with challenges such as semantic gap, insufficient knowledge dynamics and multimodal fusion bottleneck. This review aims to systematically sort out the technical framework of automatic question answering system, analyze its performance bottlenecks, and explore innovative solutions based on large language model and multimodal fusion. This paper analyzes in detail the status quo, advantages, limitations and applicable scenarios of four types of methods, such as retrieval question answering, knowledge base question answering, deep learning driven framework and multimodal question answering, and combs typical previous experimental data. The performance bottleneck is analyzed. Aiming at the illusion problem of large language model and the problem of cross-modal semantic alignment, the future research directions and technical routes are proposed, including designing hybrid architecture to integrate retrieval and generation, developing lightweight cross-modal alignment algorithms, and constructing dynamic knowledge update systems. It is hoped that the Question Answering (QA) system will develop in a more intelligent and practical direction and provide reference for subsequent research.