An approach for constructing expert yellow pages for community question answering sites. Issue 4 (13th April 2021)

• Record Type: Journal Article
• Authors: Li, Ming; Qi, Xiaoyu; Li, Ying; Lu, Xiuzhi; Wang, Li
• Other Names: Chakraborty Tanmoy guestEditor.; Bhatia Sumit guestEditor.; Caragea Cornelia guestEditor.; Moreira Fernando guestEditor.; Rocha Álvaro guestEditor.; Dubey Ashwani Kumar guestEditor.
• Is Part Of: Expert systems. Volume 38:Issue 4(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-04-13
• Subjects: artificial neural network -- community question answering -- expert yellows pages -- knowledge management; Expert systems (Computer science); 006.33
• DOI: 10.1111/exsy.12684
• Languages: English
• ISSNs: 0266-4720

Abstract

Abstract: The rapid increase in the number of community‐based question‐and‐answer services is attracting many users. Questions are posted and answered by community members. These users, who can help other users answer questions, can be considered experts. To facilitate finding a suitable expert and alleviate information overload, in this paper, expert yellow pages (EYP) for community question answering (CQA) are constructed. Considering the various lengths of texts, the biterm topic model (BTM) is used to model questions and fields of expertise. Then, two‐dimensional EYP (2DEYP), which are composed of expertise field dimensions and question dimensions, are constructed. The intersections represent the cluster of experts. The proposed 2DEYP can be expanded both laterally and vertically for a more in‐depth understanding and a more precise location. As the closer neurons represent similar topics, a novel labelling method is proposed to identify topic words for navigation. The method uses the distance between neurons as the differentiation capability. To further distinguish experts, a ranking mechanism is proposed. The experts can be ranked by integrating their expertise and activity levels. The expertise level is novel and characterized by both breadth and depth aspects. The proposed approach is evaluated via a real dataset, and the experimental results show that the proposed algorithm is feasible and performs well.