Sentiment analysis method based on Skip-gram model fusing part-of-speech and semantic information

Abstract

The invention discloses a sentiment analysis method based on a Skip-gram model fusing part-of-speech and semantic information. The method includes five steps of data preprocessing, part-of-speech information modeling, word vector representation, semantic information modeling and sentiment analysis. Data preprocessing includes filtering, word segmentation and part-of-speech labeling. Part-of-speechinformation modeling contains context-based modeling on part-of-speech information. A word vector representation module carries out vector training on the Skip-gram model fusing the part-of-speech information. The semantic information modeling module includes text representations fusing sentiment semantic information. Compared with the prior art, the method considers the part-of-speech information and the sentiment semantic information of words, fully uses the part-of-speech information of the words for helping in word vector training and the sentiment semantic prior-information for helping in learning of text vectors on the basis of modeling on the part-of-speech information of the words and fusion of the semantic information, enables the represented text vectors to more accord with linguistic features, and has good results for sentiment analysis.

Description

technical field [0001] The invention belongs to the field of natural language processing, and in particular relates to an emotion analysis method based on a Skip-gram model integrating part-of-speech and semantic information. Background technique [0002] With the development of e-commerce, sentiment analysis and mining of product review texts are of great value for researching product reputation and product recommendation. Review data has become an important data source for companies to improve product quality and service. [0003] With the great achievements of deep learning in the fields of audio, image, and video, the traditional machine learning model is fused with the neural network model in deep learning, and the traditional word vector is replaced by a low-dimensional word vector that can measure the semantic correlation between words. The one-hot vector used in the bag model has achieved superior performance compared to traditional machine learning methods in variou...

AI Technical Summary

Problems solved by technology

[0005] To sum up, most of the current research work on sentiment analysis focuses on proposing different deep neural network models for sentiment analysis. Model improvement mostly focuses on modifying the model structure to reduce model complexity. Other work mainly focuses on cross-language However, there are few sentiment analysis works that integrate word vector training with part-of-speech information and comprehensive emotional semantic information, and the use of part-of-speech information is not reasonable. Most of them choose part-of-speech information instead of integrating part-of-speech information into word vectors. Model

Images