Progressive Semantic Reinforcement for Product Feature Grouping in Twitter

  • 李 佩書

Student thesis: Master's Thesis


In recent years with the popularity of smartphones and the prevalence of social networking services people more often share their lives and comments quickly in these services than in the traditional blog websites Therefore social networking services are the valuable sources for the product feature opinion mining However generally the “quick” comments are quite short lack complete grammatical and syntactical structures; even they are full of abbreviations irregular expressions and URLs etc In addition their product feature expressions are more diverse Thus the information sparse problem will lead to more difficulty in the product feature grouping In order to solve the problems above we propose a progressive semantic reinforcement approach for product feature grouping in Twitter We combine the literal characteristics to help the semantic characteristics which are rough initially and then aggregate the related information progressively through the sub-concept group information to enhance the semantic characteristics for achieving the better grouping results Our experimental results show that our approach is better than the ones which only use one of the conditions above or neither These results prove that adding other characteristics and aggregating information progressively can be helpful for understanding the semantics in the sparse information
Date of Award2016 Dec 27
Original languageEnglish
SupervisorHung-Yu Kao (Supervisor)

Cite this