Research Of Review Text Based On Sentiment Classification And Topic Mining

With the increase of Internet penetration rate,various network platforms have begun to develop rapidly,followed by a large number of comments containing subjective emotions of users.Text sentiment analysis is one of the research hotspots in recent years.It digs out the hidden information through the sentiment classification and opinion extraction of massive text data,so as to provide different perspectives for different groups of people.From the perspective of application,this paper analyzes user reviews of different mobile phones and proposes corresponding suggestions based on the technology of sentiment classification and topic mining.It enriches the existing Chinese text mining application research in online shopping user reviews.The research ideas and methods can provide theoretical reference for text research in different industries,which has great theoretical and practical significance.First,this article uses Python crawler technology to collect user comment data of 12 mobile phones in four price ranges from the Jingdong platform.After word segmentation and stop word processing,ALBERT word embedding technology is used for text vectorization.The obtained vectorized text is trained under different deep learning sentiment classification frameworks,and the classification effects of the classification models under different frameworks are compared to find the relatively optimal sentiment classification framework.Then,the review text is excavated and visually displayed through LDA theme to fully understand the advantages and disadvantages of mobile phones at different price ranges.Based on this,corresponding suggestions are put forward from the three perspectives of people,goods,and markets,namely consumers,mobile phone developers,and e-commerce platforms.Judging from the evaluation indicators of the four sentiment classification frameworks,deep learning sentiment classification based on the Bi GRU model and the Attention mechanism has the best effect relatively,and the following two conclusions can be drawn: 1.Bi GRU model has a great advantage in capturing the context information of text when processing temporal data like text;2.Attention mechanism can make the sentence semantic expression more accurate through weight distribution.The results of the LDA theme model show that low-priced mobile phones have high volume and strong endurance,but run stuttering,low screen resolution,and unclear pictures;low-and medium-priced mobile phones have clear subdivision functions(photographs or games),but the signal is poor,and it is easy to get hot and freeze;mid-to-highpriced mobile phones have the advantages of high display,high pixels,and high sound quality,but they are prone to hot and freeze and consume fast power;the key advantage of high-priced mobile phones is smooth operation,but there is disadvantages like the green side of the screen or the accessories are not sold together.Based on this,this article puts forward the following suggestions: 1.For consumers,they should avoid blind obedience and rational consumption according to their own needs;2.For developers,while optimizing the gaps and ensuring quality control,they should devote themselves to the research and development of their own core technologies;3.For e-commerce platforms,we should continue to improve service levels and formulate insured services.