A Consideration of Evaluation Method of Sentiment Analysis on Social Listening

In recent years, companies around the world are using media information, blogs, word-of-mouth information, etc., and there is an increasing need to utilize them in corporate activities as customer voices. A sentiment analysis engine has been developed to extract sentiment from customer voices based on these needs. The sentiment analysis engine can give a good deal of sentiment information from the document by using customer voices as text data. In order to improve the accuracy, consideration of context and meaning in text data analysis is important with maintenance of a sentiment dictionary. Especially, since blog and word-of-mouth information is often described in a colloquial manner, maintenance must be performed regularly. An appropriate evaluation of sentiment analysis results is necessary to carry out appropriate maintenance. In case of considering the global strategy of the sentiment analysis engine, it is important to eliminate the language characteristics as much as possible in order to efficiently carry out the evaluation and improvement of the sentiment analysis engine. On the other hand, an evaluation that considers the national character and customs of each country is also required. This paper describes the correction method as well as the evaluation method of the sentiment dictionary which is important in sentiment analysis engine development.