Learning based Approaches for Addressing Challenges in Sentiment Analysis

Abstract

Sentiments and their implications play a vital role in our life. People appraise newlinethe opinions of others before reaching a conclusion. With the advent newlineof forums, blogs and social networking sites, there has been a considerable increase newlinein sharing of opinions and sentiments on social media. Netizens exhibit newlinedivergent views on a particular subject and many times end up agreeing to the newlinedisagreement. A substantial amount of research has been done in automated newlinetext analysis, viewpoint analysis and opinion extraction. The increase in usage newlineof social media in multilingual countries, such as India, has posed challenges newlineof accommodating multilingual text in addition to monolingual text and its newlineprocessing. The immense increase in information overload of viewpoints on newlinethe internet, calls for efficient methods to extract the useful information with newlinethe view to facilitate decision making. Opinion mining or viewpoint excavation newlinehas been treated as a classification problem which stratifies documents newlineor products as good/bad or positive/negative. People may have ambivalent newlineopinions about the topic or product. newlineInformation overload has lead to increased focus on the need for creating an newlinealternative approach for representation and selection, of text and multimedia newlinecontents. An efficient framework is needed for representing the essential parts newlineof the text so that user can decide whether he/she should read the whole text newlineor not. For such cases, text summarization is a better suited for representing newlinean opinion. newlineAlthough the automated text summarization is focused on text inputs, but newlinemultimedia information, video, images, recorded tracks, online information newlineor hypertexts can also be considered as inputs. newlineThe thesis focuses on the problem of text summarization which considers vital newlineparts of the document, extracts utilitarian information, and provides a broad newlineoverview of opinions. This saves the user from going through many documents newlineto reach a conclusion. Work done also focuses on aspects/features newlinepresent in the document and associates op