Building Place-Specific Sentiment Lexicon

With the development of various smart applications, reviewing and grading places using location-based services has become a routine task among users. The user evaluations can significantly affect not only potential visitors, but also service providers. However, too much data interrupts people's reasonable decision. In this paper, we propose a method to construct a place-specific sentiment lexicon by analyzing place reviews written by users. To this end, word2vec and association rule mining were used to extract place features and corresponding sentiment words, and sentiment word combinations were generated. Then, the probability values for each place sentiment word combination were calculated through bag-of-words logistic regression. As a result, the place sentiment lexicon for 347 sentiment word combinations were constructed. We analyzed the reviews of Italian restaurant among various place categories, and place sentiment lexicons of other kinds of places can also be constructed according to the method proposed in this study. The place-specific sentiment lexicon allows the context to be reflected in the sentiment analysis of places.