Abstract
Convolutional neural networks (CNN) and long short-term memory (LSTM) have gained wide recognition in the field of natural language processing. However, due to the pre- and post-dependence of natural language structure, relying solely on CNN to implement text categorization will ignore the contextual meaning of words and bidirectional long short-term memory (BiLSTM). The feature fusion model is divided into a multiple attention (MATT) CNN model and a bi-directional gated recurrent unit (BiGRU) model. The CNN model inputs the word vector (word vector attention, part of speech attention, position attention) that has been labeled by the attention mechanism into our multi-attention mechanism CNN model. Obtaining the influence intensity of the target keyword on the sentiment polarity of the sentence, and forming the first dimension of the sentiment classification, the BiGRU model replaces the original BiLSTM and extracts the global semantic features of the sentence level to form the second dimension of sentiment classification. Then, using PCA to reduce the dimension of the two-dimensional fusion vector, we finally obtain a classification result combining two dimensions of keywords and sentences. The experimental results show that the proposed MATT-CNN+BiGRU fusion model has 5.94% and 11.01% higher classification accuracy on the MRD and SemEval2016 datasets, respectively, than the mainstream CNN+BiLSTM method.
Highlights
Accompanying the continuous development of social networks, the role of Internet users has changed quietly from the original recipient of information to the creator of information
We found that the proposed multiple attention (MATT)-Convolutional neural networks (CNN)+bi-directional gated recurrent unit (BiGRU) model had a dominant position in the ratios of Laptop and movie review data (MRD) datasets
Comparing the proposed MATT-CNN+BiGRU model with the CNN+bidirectional long short-term memory (BiLSTM) model, we found that the fusion model of the two significantly could improve the classification accuracy compared to the variants of the single long short-term memory (LSTM) and CNN models
Summary
Accompanying the continuous development of social networks, the role of Internet users has changed quietly from the original recipient of information to the creator of information. The current research finds that the combination of the CNN network and the attention mechanism can obtain very good target-specific emotional classification results [11]. It can well solve the shortcomings of LSTM which cannot accurately indicate the importance of each word in the sentence. The proposed fusion model combines this advantage and the CNN model with a multiple-attention mechanism that is proposed to obtain the emotional polarity of keywords, which is an important dimension of the emotional classification of fusion models. Aiming at the current problems in the field of short text sentiment classification, this paper proposes a feature fusion text classification model combining CNN and BiGRU with a multi-attention mechanism based on previous research The main contributions of this paper are as follows: 1. Aiming at the current problems in the field of short text sentiment classification, this paper proposes a feature fusion text classification model combining CNN and BiGRU with a multi-attention mechanism based on previous research
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
