Bayesian estimation‐based sentiment word embedding model for sentiment analysis

Abstract

Sentiment word embedding has been extensively studied and used in sentiment analysis tasks. However, most existing models have failed to differentiate high-frequency and lowfrequency words. Accordingly, the sentiment information of low-frequency words is insufficiently captured, thus resulting in inaccurate sentiment word embedding and degradation of overall performance of sentiment analysis. A Bayesian estimation-based sentiment word embedding (BESWE) model, which aims to precisely extract the sentiment information of low-frequency words, has been proposed. In the model, a Bayesian estimator is constructed based on the co-occurrence probabilities and sentiment probabilities of words, and a novel loss function is defined for sentiment word embedding learning. The experimental results based on the sentiment lexicons and Movie Review dataset show that BESWE outperforms many state-of-the-art methods, for example, C&W, CBOW, GloVe, SE-HyRank and DLJT1, in sentiment analysis tasks, which demonstrate that Bayesian estimation can effectively capture the sentiment information of low-frequency words and integrate the sentiment information into the word embedding through the loss function. In addition, replacing the embedding of low-frequency words in the state-of-the-art methods with BESWE can significantly improve the performance of those methods in sentiment analysis tasks.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

A semantic and emotion‐based dual latent variable generation model for a dialogue system

“…The CVAE framework [28] is expressed as an end‐to‐end generation model by introducing three random variables: post x , generated response y and input conditions c , which are used for estimating latent variable z and denoting emotional tag e in emotional dialogue. CVAE can be effectively trained by using the random gradient variable Bayesian (SGVB) [29] by maximising the lower bound of variation of x under the given condition of c : $$$p(y\vert x,c)=\underset{z}{\int }p(y\vert x,c,z)p(z\vert c)dz$$$ …”

Landscape Aesthetic Value of Waterfront Green Space Based on Space–Psychology–Behavior Dimension: A Case Study along Qiantang River (Hangzhou Section)

“…Finally, in order to further analyze the characteristics and differences in users’ sensory evaluations, this study conducted word frequency analysis on the parks with emotional values greater than 1 based on the results of relative emotional values. The word frequency analysis technology was used to segment, check, and fit the evaluation of the park [ 75 , 76 ], and people’s emotional perceptions were obtained from the five aspects of vision, smell, hearing, touch, and taste, so as to analyze the results of the crowd experience of different waterfront spaces.…”

Sentiment analysis using multi neighborhood representation learning embeddings: graph convolutional neural networks

“…Che et al proposed a framework for Sent_Comp which adds a sentence compression step before sentiment analysis to address the problem that sentiment analysis depends largely on syntactic features [5] . Tang et al presented the Bayesian Estimation based Sentiment Word Embedding (BESWE) model which accurately extracts sentiment information of low-frequency words [6] . Zhu et al introduced a framework and features for multimodal sentiment analysis and cross-modal sentiment information integration [7] .…”