DEEP-LEARNED CROSS-DOMAIN SENTIMENT CLASSIFICATION USING INTEGRATED POLARITY SCORE PATTERN EMBEDDING ON TRI MODEL ATTENTION NETWORK
Parvati Kadli -, Vidyavathi B M · Indian Journal of Computer Science and Engineering · 2021
Classification of sentiments is a challenging task when it is done across multiple domains.The dissimilarity between the domains, reflected in their sentiments makes the classification tougher.This work proposes a deep learned novel architecture named Integrated Polarity Score based Pattern Embedding on Tri Model Attention (IPSPE_TMA) Network which is a model based on Bidirectional Long-short Term Memory (Bi-LSTM), Bi-Gated Recurrent Unit (GRU) and Convolutional Neural Network (CNN) with three stage novel embedding architecture and attention network.The information is extracted by incorporating attention in word and sentence level.In order to generate robust vectors, a novel polarity score based embedding is proposed and used along with Glove and Fast Text in this paper.Integration of three models with three kinds of embeddings and attention mechanism strengthens this model.This network implementation for sentiment analysis on cross domain dataset gives better performance than many of the previous works.