Design And Implementation Of The Text Sentiment Analysis System Based On Deep Learning

In recent years,due to the development of Internet technology and the improvement of people's living standards,users have increasingly higher requirements towards the using experience of system and service quality.Hebei Sibosi Innovation Technology Co.,Ltd.also has a deep foundation in the filed.The company can not appraise the customer service stuff in time because of the lack of sentiment analysis system for texts,which results in the company not being able to improve the service method individually.After the system released new features,it can not understand the user's experience in time and when the police issued a new notice or a specific event,it can not make decisions because the emotional tendencies of various companies or net users could not be known in time.Aiming at these problems,the company proposed to establish a text sentiment analysis to lay the foundation for further business development by collecting texts about the company on the analysis forum.This article takes the Tianya Forum text and the company's existing data as an example to conduct researches deeply.Through analyzing the needs and designing the overall framework of the system,a text sentiment analysis based on deep learning is developed for sentiment analysis of web texts.The main tasks are as follows:(1)Data acquisition and category labeling.We can get the text data of Tianya Forum by writing the web crawler program.We can using the BeautifullSoup to parse,extract the data in wen page and save it to the database.Then we classify the obtained data.(2)Data preprocessing.Firstly,we perform data cleaning on the obtained original text and use the word segmentation tool to segment the cleaned text.Finally,we filter the stop words.(3)Model training.It includes word vector model training and sentiment analysis model training.In order to determine the sentiment polarity of the text better,a sentiment classification based on Long Short-Term Memory neural network model is proposed,which establishes a corpus,used as the original feature vector of the classification model,by Word2 vec word vector training.Next,we further extract features by sending it to the Long Short-Term Memory neural network model and get the sentiment classification model.Thenwe use the test text to evaluate and appraise the classifier.The experiment compares the Support Vector Machine in traditional machine learning,the Recurrent Neural Network and Long Short-Term Memory neural network in deep learning.Finally,it is proved that the Long Short-Term Memory neural network model is the best for emotion classification of forum texts.(4)Based on the above research,we design and implement a text sentiment analysis system on the foundation of deep learning,which includes four modules: data acquisition and storage,data preprocessing,model training and user interaction.The classification accuracy rate reached 82%.