An enhanced semantic tree kernel for sentiment polarity classification

摘要: Sentiment analysis has gained a lot of attention in recent years, mainly due to the many practical applications it supports and growing demand for such applications. This is supported by an increasing amount availability opinionated online information, proliferation popularity social media. The majority work sentiment considers polarity word terms rather than specific senses context. However there been increased effort distinguishing between different as well their opinion-related properties. Syntactic parse trees are widely used natural language processing construct that effectively employed text classification tasks. paper proposes novel methodology extending syntactic trees, based on sense disambiguation context features. We evaluate three publicly available corpuses, employing sub-set tree kernel similarity function support vector machine. also effectiveness several lexicons. Experimental results show all our extended representations surpass baseline performance every measure across compared other state-of-the-art techniques.