Research On Fake Review Detection Method Based On Sentiment Intensity And PU Learning

In dealing with massive online product reviews,the detection of fake reviews is always an indispensable task for maintaining the healthy development of the online environment.Although great progress has been made in previous fake reviews detection work,the detection of deceptive reviews is still difficult due to the hidden and diverse characteristics of deceptive reviews.For such an issue,this paper presents a fake review detection model based on sentiment intensity and PU learning(SIPUL).Different from the previous work,in order to improve the detection efficiency of deceptive reviews,this paper innovatively proposes to use the different sentiment intensity between reviews to divide and change the dataset of fake reviews through sentiment calculation,so that the original hidden deceptive reviews can be fully displayed in the new dataset environment.The main work of this paper is as follows:(1)Aiming at the problem of mismatch between basic sentiment dictionary and fake review text information elements in Affective computing,a product review text affective computing model based on sentiment dictionary is constructed.First,use TFIDF to extract candidate seed words with high word frequency and strong sentiment intensity.Then use the K-means++algorithm to cluster the candidate seed words to determine the final seed word set.Then,the label propagation algorithm is used to expand the seed word set to complete the construction of the sentiment dictionary in the field of false reviews.Combining the degree adverb dictionary,the transition word dictionary and the negative word dictionary to form a sentiment lexicon in the field of fake review detection.Finally,based on the constructed sentiment dictionary,a variety of sentiment calculation specifications were formulated according to the rules of general sentences to complete the sentiment calculation of fake review texts.(2)Aiming at the problem that deceptive fake reviews are difficult to detect,a fake review detection model based on sentiment intensity and PU learning is constructed.First,based on the constructed review text sentiment calculation model,the sentiment value of a given product review is calculated,and the review is divided into two subsets of strong sentiment and weak sentiment according to the sentiment value.Then,a small number of reliable positive examples are selected using random strategies in the subdata sets,and a small number of reliable negative examples are extracted using SPY technology.Finally,based on the extracted positive and negative samples,a false comment detection model of the PU learning algorithm is constructed.The experimental results show that the construction of emotional vocabulary and the formulation of multiple sentiment calculation rules in the field of fake review detection can better match the sentiment calculation of reviews.The semi-supervised PU algorithm can well complete the training of the model based on a small number of labeled data sets.The introduction of sentiment intensity not only makes the model obtain a higher accuracy rate of fake review detection,but also obtains good test results in the deceptive review detection test.The research in this paper can effectively detect fake reviews in a given product area to maintain the fairness of the e-commerce market environment,and eliminate fake reviews to help users better understand the authenticity of product information and make correct purchase decisions.Figure[21]Table[23]Reference[73]...