On-site Impression Grasping System Using SNS Location Information and Sentiment Analysis

Abstract

In the recent years, information from such as Social Networking Service (SNS) is overflowing. It draws a great attention from research community in efficiently collecting and analyzing what is being shared in real time. Extracting topics in SNS and analyzing the emotional expression related to those topics are one of the means to know the trends of social interest. In the conventional sentiment analysis, the sentiment of a sentence is estimated on a word-by-word basis, which tends to give undesired results. In many cases, such as negative auxiliary verbs, adverbs, and adjectives related to emotional words are ignored in configurating the emotional expressions. In this research, a method of natural language processing (NLP) in sentiment analysis is performed phrase by phrase by combining the results of active text analysis with the location information. We propose a method to grasp the social impression on an event by improving the phrase level sentiment analysis combining to the location information.