CAJ | 학술논문

以 LDA( Latent Dirichlet Allocation)为代表的主题模型自提出以来就受到了广泛关注，并且随着研究的不断深入产生了大量有代表性的研究成果。为了满足用户多样化的检索需求，本文立足于科技情报分析工作，从融合单一外部特征和同时融合多个外部特征两个角度综述了一些典型主题模型，并对其进行了客观分析和评价。所涉及的外部特征主要包括：科研人员、时间、参考文献、母体文献以及合著关系等。经本文深入分析发现，目前融合单一外部特征的主题模型研究已经趋于成熟，而同时融合多个外部特征的主题模型研究仍处于起步阶段。缘于，模型融合的外部特征越多，能反映出的文献信息种类就越多，从而能更好地满足用户多样化的实际需求。因此，本文认为同时融合多个外部特征的主题模型研究将是科技情报分析未来的发展方向之一，应加大研发力度。
이 LDA( Latent Dirichlet Allocation)위대표적주제모형자제출이래취수도료엄범관주，병차수착연구적불단심입산생료대량유대표성적연구성과。위료만족용호다양화적검색수구，본문립족우과기정보분석공작，종융합단일외부특정화동시융합다개외부특정량개각도종술료일사전형주제모형，병대기진행료객관분석화평개。소섭급적외부특정주요포괄：과연인원、시간、삼고문헌、모체문헌이급합저관계등。경본문심입분석발현，목전융합단일외부특정적주제모형연구이경추우성숙，이동시융합다개외부특정적주제모형연구잉처우기보계단。연우，모형융합적외부특정월다，능반영출적문헌신식충류취월다，종이능경호지만족용호다양화적실제수구。인차，본문인위동시융합다개외부특정적주제모형연구장시과기정보분석미래적발전방향지일，응가대연발력도。
Topic models represented by LDA ( Latent Dirichlet Allocation) have received extensive attention since it was proposed,and produced a lot of research results with intensive research. From the perspective of information analysis, this paper reviews those topic models integrating single or multiple extra-features of scientific and technical literatures from some aspects such as authors,time,references,publication venues,coauthor relationship and so on and compares them objectively for the users’ various requirements. This paper finds by thorough study that the research on topic models integrating single extra-feature has matured and integrating multiple extra-features counterpart has just started. Since,topic models integrating more extra-features can reveal more information,they can meet the users,various demands better. Hence,this paper thinks that the research on topic models integrating multiple extra-features will be one of the future trends and need to be paid more attention.