HDAS: A Hybrid Deep Learning Approach for Aspect based Sentiment Analysis

Abstract

The observation and analysis of a person’s opinion on a specific context out of the reviews given is called sentiment analysis.  One of this sentiment analysis is Aspect mining in which sentiment analysis performed on not only the entire product but individual features of the product/ context. The task of aspect mining is really challenging as the features or properties of the product may implicitly present in the text as a feedback given by the user and calculating polarity relative to corresponding properties present in the text is also difficult. From the past few years, several authors proposeda vast number of models to handle these issues. Reverse Document frequency, Term frequency, Lexicon-based approaches are a few of the models from these and reverse document frequency are some of the suggested approaches by the authors. In this article we aim to propose an effective method to analyze the sentiments at aspect level by using Semantic feature extraction, further by using Word to Vector (Word2vec) transform the extracted aspects and finally applying a hybrid deep learning model that comprises of both Support Vector Machines and Convolutional Models for the mining of opinion. SemEval 2016 dataset is used to implement and test the proposed model. Finally, the conclusions were drawn.