Survey on Deep Learning Approaches for Aspect Level Opinion Mining

Abstract

The the task of Aspect-based opinion mining (AbOM) is an emeraging research area, where aspects are mined, the corresponding opinion are scrutinized and sentiments are continuously changed, is gaining increased attention with growing feedback of clients and community across various social media streams. The gigantic improvements of deep learning (DL) techniques in natural language processing (NLP) tasks motivated research community to introduce a novel DL models and for AbSA, each investigate a diverse research points from different perspective, that cope with imminent problems and composite circumstances of AbOM. Consequently, in this survey paper, we concentrate on the limitations of the current studies and challenges relevant to mining of various aspects and their pertinent opinion, interrelationship delineations among different aspects, interactions, dependencies and contextual-semantic associations among various entities for enhanced opinion precision, and estimation of the automaticity of opinion polrity development. A laborious investigation of the later advancement is discussed depending on their contribution in the direction of spotlighting and alleviating the shortcomings related to Aspect Extraction (AE), AbOM, opinion progression (OP). The reported performance for each scrutinized study of Aspect Extraction and Aspect opinion Analysis is also given, revealing the numeriacal evaluation of the presented approach. Future research trends are introduce and delibrated by critically analysing the existing recent approaches, that will be supportive for researchers and advantageous for refining aspect based opinion classification.