Multilingual Sentiment Analysis for Web Text Based on Word to Word Translation
Keita Fujihira, Noriko Horibe
Abstract
People's sentiments are known to have a large impact on changes in stock prices, products sales, and trends. Since web users generally state their opinion in various languages, it is important to develop a method of multilingual sentiment analysis for web texts. In this research, we design a multilingual sentiment analysis method based on word to word translation using a sentiment dictionary in arbitrary native language. This method consists of three phases: morphological analysis of text, a sentiment extraction of each word with sentiment dictionary, and a sentiment extraction of text based on words sentiments. We conduct a sentiment classification experiment for tweets in English, German, French, and Spanish. In the experiment, we evaluate our classifier's performance by comparing the classifier with the other previous classifiers based on the evaluation standards "Accuracy", "Precision", "Recall", and "F1 score". The experimental results show that our classifier has an applicability to sentiment analysis for multilingual, because our classifier's performance is independent of the differences languages.