Research And Application Of Aspect-oriented Sentiment Analysis And Evaluation System Of Home Appliance Product Reviews

With the continuous research and development of the accuracy and depth of text analysis in the field of natural processing and the rapid accumulation of product evaluation information in Internet e-commerce platforms,it is very urgent to dig out more valuable and more accurate information.Therefore,the analysis of text sentiment tendency has developed from the text level to the current aspect level text sentiment analysis.This paper conducts the mining of aspect-level sentiment information on the online product evaluation of home appliances and proposes a sentence-level sentiment analysis method based on a probability generation model and an LSTM aspect-level analysis method based on an attention mechanism.The analysis model is used as technical support to construct an analysis system for the evaluation of home appliances.The research in this paper is mainly focused on the following four aspects:(1)Propose a sentence-level sentiment analysis method based on the probability generation model.In response to the current major e-commerce platforms that have not set strict standardized restrictions on evaluation content,the evaluation content does not match the evaluation star rating.This paper proposes to use a probabilistic generation model to build "evaluation-topic-words" The mapping relationship is used to analyze the emotional tendency of the product evaluation data,and then identify the data in the data set that does not match the star rating.(2)An LSTM aspect-level sentiment analysis method based on the attention mechanism is proposed.Aiming at the problem that the amount of Chinese aspect-level data is small,and the coverage area is not wide,a processing method is proposed to construct an aspect-level analysis model after first using the model to learn the information in the document-level data.In this paper,this model is applied in the field of sentiment analysis of Chinese evaluation data of home appliances,and a comparative experiment is carried out with the LSTM model that does not incorporate the attention mechanism.(3)Designed the construction method of the commodity evaluation system.This paper uses the proposed model to construct a home appliance product evaluation analysis system.To display the evaluation results of various aspects of the product according to the distribution of evaluation objects in the product evaluation data,a method based on semantic fusion is proposed to construct a product evaluation system.This article applies the evaluation system to the display part of the system analysis results and plays a role in determining which aspects of the product analysis results are displayed.(4)Designed and implemented a prototype system for intelligent evaluation and analysis of home appliances.Based on the sentiment analysis model proposed in this paper,the technical framework of separation of front and back ends is adopted to provide consumers and enterprise manufacturers with fast and accurate data mining tools for product evaluation.The system is divided into data management module,analysis result query module and task management module.The data management module uniformly manages the product evaluation data and updates it regularly;the analysis result query module performs sentence-level and aspect-level sentimental analysis on the product evaluation data,and can query product analysis results,which compares data and analysis results comparison of different models;the task management module centrally manages the analysis tasks and task status performed by the system.