Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LB-KBQA: Large-language-model and BERT based Knowledge-Based Question and Answering System

Yan Zhao, Zhongyun Li, Yushan Pan, Jiaxing Wang, Yihong Wang

TL;DR

This paper tackles unseen intents in knowledge-based question answering (KBQA) by introducing LB-KBQA, a system that fuses large language models (LLMs) with BERT-based representations to capture diverse user queries. The architecture comprises five modules—language preprocessing, intent recognition, response generation, adaptive learning, and query library extension—designed to detect newly appearing intents and update knowledge representations through user feedback. In a finance-domain case study using the Tushare dataset, the approach achieves high accuracy (0.90) and demonstrates that combining rule-based, BERT-based semantic embedding, and LLM fallback effectively handles linguistic diversity and unseen intents, with adaptive learning further refining intent tracking. The work provides practical methodology and empirical evidence for deploying Generative AI-enabled KBQA in industry, enabling dynamic knowledge updates and scalable domain-specific QA. Key contributions include identifying unseen intents (divided into diverse representations and newly appeared intents), integrating prompt-learning-based LLMs for intent generation, and validating the approach through ablation studies that highlight the importance of BERT representations and adaptive learning for robust KBQA in the financial sector.

Abstract

Generative Artificial Intelligence (AI), because of its emergent abilities, has empowered various fields, one typical of which is large language models (LLMs). One of the typical application fields of Generative AI is large language models (LLMs), and the natural language understanding capability of LLM is dramatically improved when compared with conventional AI-based methods. The natural language understanding capability has always been a barrier to the intent recognition performance of the Knowledge-Based-Question-and-Answer (KBQA) system, which arises from linguistic diversity and the newly appeared intent. Conventional AI-based methods for intent recognition can be divided into semantic parsing-based and model-based approaches. However, both of the methods suffer from limited resources in intent recognition. To address this issue, we propose a novel KBQA system based on a Large Language Model(LLM) and BERT (LB-KBQA). With the help of generative AI, our proposed method could detect newly appeared intent and acquire new knowledge. In experiments on financial domain question answering, our model has demonstrated superior effectiveness.

LB-KBQA: Large-language-model and BERT based Knowledge-Based Question and Answering System

TL;DR

This paper tackles unseen intents in knowledge-based question answering (KBQA) by introducing LB-KBQA, a system that fuses large language models (LLMs) with BERT-based representations to capture diverse user queries. The architecture comprises five modules—language preprocessing, intent recognition, response generation, adaptive learning, and query library extension—designed to detect newly appearing intents and update knowledge representations through user feedback. In a finance-domain case study using the Tushare dataset, the approach achieves high accuracy (0.90) and demonstrates that combining rule-based, BERT-based semantic embedding, and LLM fallback effectively handles linguistic diversity and unseen intents, with adaptive learning further refining intent tracking. The work provides practical methodology and empirical evidence for deploying Generative AI-enabled KBQA in industry, enabling dynamic knowledge updates and scalable domain-specific QA. Key contributions include identifying unseen intents (divided into diverse representations and newly appeared intents), integrating prompt-learning-based LLMs for intent generation, and validating the approach through ablation studies that highlight the importance of BERT representations and adaptive learning for robust KBQA in the financial sector.

Abstract

Generative Artificial Intelligence (AI), because of its emergent abilities, has empowered various fields, one typical of which is large language models (LLMs). One of the typical application fields of Generative AI is large language models (LLMs), and the natural language understanding capability of LLM is dramatically improved when compared with conventional AI-based methods. The natural language understanding capability has always been a barrier to the intent recognition performance of the Knowledge-Based-Question-and-Answer (KBQA) system, which arises from linguistic diversity and the newly appeared intent. Conventional AI-based methods for intent recognition can be divided into semantic parsing-based and model-based approaches. However, both of the methods suffer from limited resources in intent recognition. To address this issue, we propose a novel KBQA system based on a Large Language Model(LLM) and BERT (LB-KBQA). With the help of generative AI, our proposed method could detect newly appeared intent and acquire new knowledge. In experiments on financial domain question answering, our model has demonstrated superior effectiveness.
Paper Structure (11 sections, 1 figure, 1 table)

This paper contains 11 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. System Design
  3. Language Preprocessing Module
  4. Intent Recognition Module
  5. Response Generation Module
  6. Adopted Learning Module
  7. Query Library Extension Module
  8. Case Study
  9. Experiment Setting
  10. Experiment Result
  11. Conclusion

Figures (1)

  • Figure 1: The LB-KBQA system is comprised of five distinct components: 1) the language preprocessing module, 2) the intent recognition module, 3) the response generation module, 4) the adaptive learning module, and 5) the query library extension module.