SUMMARIZATION USING NTC APPROACH BASED ON KEYWORD EXTRACTION FOR DISCUSSION FORUMS

Abstract

Internet has become a ubiquitous medium of communication, be it through any social networking websites like Facebook, Twitter or any discussion forums like Yahoo Answers, Quora, Stack Overflow. One can participate in any kind of discussion ranging from politics, education, spirituality, philosophy, science and geography to medicine and many more. Often, most of the discussion forums are loaded up with data. Hence, when a new user wants to know the public opinion, it is impossible for him/her to go through all the tens or hundreds of threads or comments under a particular thematic discussion. The problem here is - we are buried in data but we starve for information. So, to solve this problem, we are proposing a novel approach called Discussion Summarization which is aimed at presenting the user with the most relevant summary containing all the important points of the discussion. This allows the user to easily and quickly grasp and catch up on the on-going conversation in a discussion thread. The summary generated follows CRS approach (Clustering and Ranking and Score calculation for each sentence).The Cluster based Summarization technique is coupled with Nested Thematic Clustering (NTC) and Corpus Based Semantic Similarity (CBSS) approaches. The summary produced is the set of top-ranked sentences (of high scores). Results have shown that a completely unbiased summary with the multidimensionality of comments is generated.