Multi-lingual Sentimental Analysis using Machine Learning Algorithm

Sentimental Analysis or Opinion Mining is increasing in today’s world. As every industry, require the opinion or review from the end-user or consumers to improve the quality of their product or service. The previous research work has been done on sentimental analysis is only limited to English data analysis and secondly the on the enhancement of accuracy by using different algorithm. In this research we overcome both the issues by developing a language-independent system using Google translator API and proposed a solution with Stanford NLP which the modeling of the English language. The simulation of the proposed work has been done in two stages. First, all the tweets are translated into a single language. English is used as the target language here. To make the research database broader, we have used the Live Google translation API that will convert all the tweets into the English language. Then, emoticons, slang language, misspellings, email id, URLs, etc. are forced to preprocessing before feature extraction. In the second phase, the stop words, which do not contribute towards the sentiment of the tweets, are removed, and the tweets are converted into feature vectors. This feature vector is then used in the classification algorithm. The results are compared using two classification techniques, i.e., Naïve Bayes classification and RNN.

Keywords: Opinion mining, Translation, Naïve Bayes, RNN.