Quality Features for Summarizing Text Forum Threads by Selecting Quality Replies

摘要

Text Forum Threads contain a huge volume of user-generated content derived from the discussion and information exchange among users who have similar interests. Often, some of the replies in a thread are completely off-topic which changes the discussion's direction. This phenomenon impacts negatively on the user's desire to continue with the discussion hence, a user might be interested in reading a few selected replies that provide a brief summary of the discussion topic. This paper aims at selecting quality replies about a topic raised in the initial-post which provide a short summary. We present a detailed analysis of the text forum threads structure based on a set of quality features for the forum summarization. Moreover, crowdsourcing platforms were used for judging the quality of the replies. Therefore, we have performed a text forum threads summarization based on replies weights and human judgment. TripAdvisor dataset has been used, therefore, the system summary helps the traveler in planning a journey. The experimental results conducted showed that the proposed approach can improve the performance of the text forum threads summarization based on forum quality features and crowdsourcing.