Research On Emotional Analysis Of Commodity Comments

In the era of Internet 2.0,a large number of subjective texts will be produced on the Internet every day.These texts are of profound value.The most representative of them are social networks,various forums,and various websites that produce review data,which converge on the real ideas of the netizens,rather than an objective description of the news.Deep excavation of these huge amounts of content can bring unexpected results.This article selects the commodity comment to carry on the emotional analysis,so that people can quickly and simply understand the main advantages and disadvantages of the goods,make the corresponding judgment and improve it according to their own concerns.In this paper,we use the Scrapy framework to capture the comment text of millet mix2 mobile phone in Jingdong mall as the source data of the experiment.After the process of data cleaning,Chinese word segmentation,disuse words,text to quantization and the establishment of a dictionary,the review text is preprocessed,and the data that can be directly input into the neural network is generated.The process of building dictionaries includes three steps:establishing index labels,sequence completion or truncation,and transformation to matrix.In the narrow sense of sentiment analysis and classification,this paper adopts a lot of preprocessing,and uses Word2Vec shallow neural network model to quantify the text in advance.Then the vectored model is entered into the embedded layer of the first layer of the LSTM depth learning model,and then the commendatory and derogatory text of the review text is judged again after training.The word vector model generated by the Word2Vec model is input into the embedded layer of the LSTM model,which is equivalent to the two direction quantization process.This method effectively improves the accuracy of text classification to 89.6%.The overall review text mining work,this article uses K-means clustering short comments,auxiliary word frequency statistics,Word2Vec keyword extraction and artificial generalization,and other means to form 8 display labels.They are all noun +adjectives,followed by all the times mentioned in this aspect.Red in the label is commendatory and green is a derogatory meaning.Clicks can check all the corresponding comments,and the corresponding short comments are red.This paper effectively improves the accuracy of Chinese text emotion analysis,and constructs a commodity review emotional analysis system,which shows the advantages and disadvantages of the goods,speeds up people's understanding of the text,and avoids the hard work of reading mass text.Consumers can make purchase decisions quickly,and businesses can make improvements to the most criticised areas,increasing sales and producing direct economic value.