Methods On Product Reliability Analysis Based On Online Reviews And Its Applications

With the popularization of the Internet,online consumption has gradually occupied the mainstream of consumption in life,and the massive amount of online reviews generated on e-commerce platforms and forums has become an effective way for companies to understand the attitude of the market and consumers.The use of online reviews to tap consumer product attitudes,analyze the reliability of product features,and provide guidance and basis for companies to improve products and innovative solutions.This article takes the web commentary as the research object,and uses the methods of review usefulness classification,product feature extra ction,part-of-speech dependency relationship rules,and viewpoint extraction to achieve consumer point of view and product reliability analysis from online reviews.The main research work of this article is as follows:(1)A classification model for the usefulness of product reviews was constructed.First,manual annotation of the review text is performed to achieve manual classification of the sample,and the data classification sample is divided into a training set and a test set.Then,a selection scheme for classification characteristics was developed.The required features and the weights of these features were obtained using the chi-square statistics and the TF-IDF algorithm,respectively.Finally,according to the selection of features and the training of data,the classification of the usefulness of product reviews is achieved,and comments that contribute to product reliability analysis are obtained.(2)Put forward an improved method for extracting product feature words.Firstly,using the conditional random field model to train through data learning,the model was established,and a standard product feature word dictionary was constructed.When the model is labelled,the word similarity calculation method is used to invoke the standard product feature word dictionary and unlabeled words that may be feature words.If the conditions are accidentally annotated with the airport,the label result is changed to better extract feature words.(3)The dependence relationship between words and words is disc overed through the dependency tree,which proposes the advantages of the part-of-speech dependency relationship for rules and rulemaking: it can more easily achieve the extraction of emotional words,negative words,and degree adverbs.On the basis of the feature word extraction method,the method of extracting opinion word pairs is proposed based on the relationship between the part-of-speech dependency rules and rules.(4)A product reliability analysis method was proposed.Using the previous score intensity processing of affective words,degree adverbs,and negative words,a product reliability analysis method suitable for this paper was proposed.The confidence interval was calculated based on the normal distribution score samples,and the reliability of product features was analyzed.