面向Web的复杂问答 （CQA）系统由于Web上可用信息的快速增长而变得越来越重要。本论文提出了一种新颖的方法，通过整合实体链接、知识库和搜索引擎技术来解决回答复杂问题所面临的挑战。我们发明了一种热插拔模块化的实体链接（HOSMEL）方法，并将其集到一个多跳问题回答系统中。我们所提出的开放领域问题回答（ODQA）系统整合了知识库和搜索引擎，为复杂问题回答提供了新的解决方案。此外，我们基于ODQA的经验为知识支撑的对话系统开发了一个有监督训练微调的大型语言模型（LLM），分别使用了自动评估和人工评估对其有效性进行评估。实验结果突显了系统的优点并揭示了现阶段依然存在的一些局限性。本研究为面向Web的复杂问答系统的发展做出了贡献，并为应对与基于Web的信息检索和自然语言理解相关的挑战和机遇的未来工作奠定了基础。

Web-oriented Complex Question Answering (CQA) systems have become increasingly important due to the rapid growth of information available on the web. This thesis presents a novel approach to address the challenges associated with answering complex questions by integrating entity linking, knowledge base, and search engine techniques. We introduce a hot-swappable modularized entity linking (HOSMEL) approach and integrate it as a multi-hop question answering (QA) framework. The proposed open-domain question answering (ODQA) system merges knowledge-base and search engine, providing a comprehensive solution to complex question answering.Furthermore, we develop a supervised fine-tuned large language model (LLM) for knowledge-grounded dialog systems and evaluate its effectiveness using automatic and human evaluations. The experimental results highlight the strengths and limitations of the system.This research contributes to the advancement of Web-oriented CQA systems and provides a foundation for future work in addressing the challenges and opportunities associated with web-based information retrieval and natural language understanding.