Development of hybrid multilayer deep neural architecture for aspect based sentiment analysis

Abstract

newline Sentiment analysis for user reviews has received substantial heed in newlinerecent years. Many deep neural models are extensively employed for Natural newlineLanguage Processing (NLP) applications. The primary function of the deep newlineneural model is to decode public reviews subjective to specific products, newlineservices, and other social activities. However, the increasing data reviews on newlinevarious aspects limit the model performance in sentiment analysis tasks. newlineTherefore, recent research focuses on developing new deep neural models for newlinesentiment analysis tasks. Long Short-Term Memory (LSTM) and newlineConvolutional Neural Networks (CNN) are popular deep neural models newlineemployed to improve sentiment classification accuracy. newlineTraditional sentiment analysis is a coarse-grained technique that newlinemainly involves classifying the overall sentiment polarity of the input newlinesentence. Such sentence-based sentiment classification assumes the newlinesentiments are unified and consistent, which is often not ideal. Alternatively, newlinefine-grained techniques analyze the sentiments based on the given specific newlinetarget or aspects, generally termed as aspect based sentiment analysis newline(ABSA). Therefore, aspects in a sentence have drawn greater attention from newlinethe research community for its more realistic assumption that sentiment is newlinedependent on a particular set of aspects or target entities. ABSA involves newlineaspect extraction, categorization, and polarity classification based on the newlinesentiment associated with the identified aspect. Based on the polarity score newlineassociated with the aspect emotions the sentiments in the dataset are classified newlineas positive, negative, and neutral.