Extracting User's Food Preferences by Sentiment Analysis

Abstract

With the growth and development of websites and social networks, the number of user comments on these platforms has grown significantly. These comments contain rich information, which can be analyzed to discover individuals' preferences in various areas, including food. Extracting individuals' preferences can be useful for many applications of the Internet of Things paradigm. This paper proposes a method for extracting individuals' food preferences from their comments. The method includes extracting foods names from individual comments, clustering them, and performing sentiment analysis for each name. Comments on the Trip Advisor website have been used in experiments. In this regard, 100 users have been chosen and their comments from January to September of 2018 have been collected. Data from the first six months has been used for training the proposed method, while the data from the last 3 months has been used for testing. The results indicate the high precision of the proposed method in extracting users' food preferences.