CAJ | 학술논문

开放域问答是当前自然语言处理和信息检索领域的研究热点，作为开放域问答系统的重要组成部分，问题分类可以缩小答案的搜索空间并决定答案的选择策略。近年来，基于机器学习的问题分类技术受到广泛的关注，相关研究表明问题分类的准确性直接影响问答系统的整体性能。本文从分类体系与数据集、特征提取、分类器设计、性能评测等层面，总结了问题分类技术近年的主要研究成果。重点分析了各种基于监督学习的问题分类方法的特点和不足，讨论了核方法、半监督学习、主动学习、迁移学习等在问题分类中的应用，同时对问题分类技术未来研究动向进行了展望。
개방역문답시당전자연어언처리화신식검색영역적연구열점，작위개방역문답계통적중요조성부분，문제분류가이축소답안적수색공간병결정답안적선택책략。근년래，기우궤기학습적문제분류기술수도엄범적관주，상관연구표명문제분류적준학성직접영향문답계통적정체성능。본문종분류체계여수거집、특정제취、분류기설계、성능평측등층면，총결료문제분류기술근년적주요연구성과。중점분석료각충기우감독학습적문제분류방법적특점화불족，토론료핵방법、반감독학습、주동학습、천이학습등재문제분류중적응용，동시대문제분류기술미래연구동향진행료전망。
Open-domain question answering is becoming a hot topic in the fields of natural language processing and informa-tion retrieval.Question classification,as an important component of question answering,has shown its significant influence on the overall performance of question answering systems.It can help reduce the search space and choose the exact search strategy to find answers.In this paper,we present a through overview of the state-of-the-art approaches to question classification,in terms of catego-ry/dataset,feature extraction,classification methods and performance metrics.Firstly,we give a detailed analysis of the supervised learning based question classification approaches.Then,we introduce some related work on question classification,such as kernel methods,semi-supervised learning methods,active learning and transfer learning methods,and so on.Finally,we give some possible research directions on question classification.