An intelligent system for Aspect based opinion mining subtasks Using machine learning techniques and graph analysis

Abstract

Opinion mining (also called sentiment analysis) is a type of natural newlinelanguage processing for computing people s opinions and emotions. It detects newlineopinions from structured, semi-structured, and unstructured social media newlinecontents at different levels, such as the document, word, sentence, and aspect newlinelevels. In all these levels except aspect, opinion mining identifies the overall newlinesubjectivity or sentiment polarities. An aspect level is described as an attribute newlineor a part of an entity. It exactly describes people s likes and dislikes in social newlinemedia content. Especially, online users express their opinions on a large newlinenumber of aspects such as food, service, restaurant, price, and quality. newlineTherefore, aspect-based sentiment analysis (ABSA) trends exist in aspectterm newlinecategorization, joint aspect-term polarity extraction, implicit ABSA, newlinemodeling inter-aspect relations, transfer learning in ABSA, and the need for newlinelarge datasets. Moreover, aspect identification is the hardest task in the newlineABSA. However, the sentiment detection task becomes relatively easier due newlineto three classes such as positive, negative, and neutral. Therefore, an newlineintelligent system for aspect-based opinion mining subtasks using machine newlinelearning and graph analysis is presented in this thesis work. The main newlineobjective of this thesis is to present four important tasks such as aspect-based newlineopining mining and ranking, co-occurrence network-based influential product newlinefeature identification, aspect category detection (ACD) with class imbalance, newlineand multilabel aspect category and sentiment polarity detection (SPD). newline