optimization of sentiment analysis of multilingual text on social network using machine learning

Abstract

In today s era, weibo and social media are the most popularly used platform for online communication. Nonetheless, Twitter is one of the social media which is being used with really a huge number of people for reflecting their opinions or beliefs about a particular topic or a product. This type of online communication has brought perfect opportunity for the businesses for stalking and overseeing their esteem; politicians to know about their reputation among the public so that the necessary action can be taken for further enhancement, as there is always a room for achieving perfection. newlineAn extensive domain of techniques for sentiment analysis has been developed recently. But most of these developed techniques reflect the sentiments on the basis of much affected words. For example, good, bad, hot, depressed etc. i.e. the semantic nature of the root words. Although, this method of identifying sentiments didn t work accurate as they do not consider the semantic meaning of the word in context of the document in which it is used. Moreover, not much of the work has been done on the multilingualism of the data. newlineThis thesis basically exposed the above mentioned problem of sentiment analysis on twitter data. Various machine learning techniques have been used for analyzing sentiments in respect to semantic as well as lexicon nature of the election tweets. The multilingual election tweets have been gathered, preprocessing is being done, then classified according to different parties like AAP, BJP, Congress, BSP, NCP. Thus, a system has been developed to recommend to the users that which party to vote on the basis of the pros and cons of the history of different parties, it can be done on any area as tweet collection application can access tweets of any area. This thesis also focuses on sarcasm detection. newlineIn this thesis, machine learning approach have been proposed and verified for uprooting the people s opinion for different parties newline