Multiclass Intent Analysis: Beyond the Conventional Polarities
Hafiz Nadeem Khan, Ayush Srivastav, Amit Kumar Mishra
Abstract
The sentiment polarity for an element determines the orientation of the conveyed feeling, namely, whether the text communicates the user's positive, negative, or neutral sentiment toward the object in question. Sentiment analysis is a crucial task in natural language processing, with numerous real-world applications. Most sentiment analysis attempts to predict whether a given text is positive or negative. One of the most efficient and robust methods applied in the domain of sentiment analysis is deep learning. The recent advancements in social networking sites and their growing impact on the behavior of individuals have led to many pieces of research in the field of study of such networking platforms. Multiclass sentiment analysis identifies the exact sentiment delivered by the tweets instead of merely generalizing those sentiments as positive, negative, and neutral. In the case of online buying, most clients resort to the reviews about the same product they want to buy that are posted on the website. This work incorporates understanding the real-time situations of different customers' reviews about several products purchased through online shopping platforms thereafter determining whether their review is an Appreciation, Complaint, Enquiry, or Non-Intent. This paper includes Max-Vote Ensemble Technique, comprising three deep learning models: Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and a combined model of CNN and LSTM. These models predict the review’s sentiment and strive to classify multiple polarities breaking through the three conventional polarity categories.