Recommendation of experts in community question answering using tag relationship / Anitha Anandhan

Anitha , Anandhan (2021) Recommendation of experts in community question answering using tag relationship / Anitha Anandhan. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (187Kb)
	PDF (Thesis PhD) Download (2078Kb)

Abstract

Community Question Answering (CQA) systems are discussion platforms for sharing knowledge rapidly through question and answer posts on social media. However, there was dissatisfaction due to slow response and low-quality answers from the crowded information. Most of the existing studies focused on expert recommendation based on available information. Still, the challenge is to find the specific domain-related experts based on user preferences for their questions. Experts are the users who post high-quality answers based on tag metadata in CQA. This condition requires experts to respond to the inquiries posted on community-based websites. Hence, this study aims to find the experts using related archive posts, users, score and tag metadata in CQA. Most CQA questions are posted with multiple tags. The co-occurrence of tags is essential for recommendation purposes as it shows the user’s interest in multiple domains, leading to high-quality answers and reducing waiting time. This study proposed the Tag Relationship Expert Recommendation (TRER) method to create the user profile by extracting the user’s interest. Similar posts are retrieved from the archive and used to rank and recommend experts for the selected domain using the input question tokens. The implicit tag relationship has been integrated into the proposed method to predict multiple domain and users relatedness. A similar user’s tag-to-tag relationship based on user preferences is utilised to recommend the specific and relevant domain experts for the input questions. Findings depict that TRER improves the performance of expert recommendations topredict the relevant and similar interest users based on the input questions by incorporating Question Answer (QA) space. Moreover, the tag relationship of the user effectively helps to find the experts with specific domain knowledge to answer the input questions posted by technical and academic professionals. The findings indicate that the TRER method outperformed baseline methods.

Actions (For repository staff only : Login required)