Scholarlink

<p>The field of AI-driven question answering has significantly evolved with advancements in large language models (LLMs), ranging from transformer-based models like BERT, which excels in understanding and retrieving text, to generative models such as ChatGPT and T5, which generate contextual responses. Within professional domains, AI-driven Q&amp;A systems enable machines to retrieve and generate precise answers by leveraging ML, natural language processing (NLP), and knowledge-based approaches. This survey aims to provide a structured analysis of AI-driven Q&amp;A systems, answering key research questions: (1) How are Q&amp;A systems categorized based on taxonomy, domain, and architecture? (2) What are the recent advancements, particularly in generative AI and knowledge-based approaches? (3) What are the key challenges, such as accuracy, hallucination in generative models, and ethical concerns? A well-defined taxonomy is introduced, classifying Q&amp;A systems into monolingual, cross-lingual, and multilingual frameworks while differentiating between open-domain and closed-domain approaches. This study extends previous surveys by systematically analyzing the impact of retrieval-based, hybrid, and generative models on Q&amp;A performance. Additionally, the survey explores critical challenges, including bias in datasets, reasoning limitations, latency in real-time applications, and the ethical concerns surrounding AI-generated content. Future research directions include enhancing model robustness, mitigating biases, improving explainability, and optimizing generative models for domain-specific tasks. By addressing these challenges and opportunities, this study provides a comprehensive evaluation of state-of-the-art Q&amp;A techniques, their limitations, and potential improvements.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A comprehensive review of AI-driven Q&A systems with taxonomy, prospects, and challenges

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Zahrah Albassami,
  • Abdulmohsen Algarni,
  • Ayman Qahmash,
  • Zulfiqar Ahmad

摘要

The field of AI-driven question answering has significantly evolved with advancements in large language models (LLMs), ranging from transformer-based models like BERT, which excels in understanding and retrieving text, to generative models such as ChatGPT and T5, which generate contextual responses. Within professional domains, AI-driven Q&A systems enable machines to retrieve and generate precise answers by leveraging ML, natural language processing (NLP), and knowledge-based approaches. This survey aims to provide a structured analysis of AI-driven Q&A systems, answering key research questions: (1) How are Q&A systems categorized based on taxonomy, domain, and architecture? (2) What are the recent advancements, particularly in generative AI and knowledge-based approaches? (3) What are the key challenges, such as accuracy, hallucination in generative models, and ethical concerns? A well-defined taxonomy is introduced, classifying Q&A systems into monolingual, cross-lingual, and multilingual frameworks while differentiating between open-domain and closed-domain approaches. This study extends previous surveys by systematically analyzing the impact of retrieval-based, hybrid, and generative models on Q&A performance. Additionally, the survey explores critical challenges, including bias in datasets, reasoning limitations, latency in real-time applications, and the ethical concerns surrounding AI-generated content. Future research directions include enhancing model robustness, mitigating biases, improving explainability, and optimizing generative models for domain-specific tasks. By addressing these challenges and opportunities, this study provides a comprehensive evaluation of state-of-the-art Q&A techniques, their limitations, and potential improvements.