Sentiment Classification using Attention based Gated-CNN with Deep Recurrent Neural Model

Abstract

Sentiment analysis received a lot of attention recently due to its potential use in business intelligence. 4 Understanding variable length sentences to extract the sentimental context is the main challenge of this concept. Our 5 proposed models are moderations of a deep neural model named comprehensive attention recurrent model [5]. A new 6 layer of attention mechanism and replacement of LSTM with gated-CNN have been introduced to make learning of CA 7 model [5] faster and efficient. IMDB movie review sentiment-labelled dataset has been used in our experiments. Our 8 paper solely focuses on the comparison of performances among proposed and inspired models. Experimental results 9 imply that accuracy and precision of our proposed models are better compared to the state-of-the-art CA model. 10