Method and apparatus for detecting abnormal product reviews online

摘要

An embodiment of the present invention detects abnormal product reviews online that automatically detects maliciously written abnormal review candidates to improve the quality of product reviews that have an important influence on consumers' product purchases in online-based shopping malls and delivery platforms, etc. To a method and an apparatus, the method for detecting abnormal product reviews on-line according to an aspect of the present invention comprises the steps of: (a) obtaining an actual user score of a specific product review; (b) extracting the text of the specific product review and deriving a sentiment vector for the extracted text through a natural language processing (NLP) sentiment analysis model; (c) predicting a user score corresponding to the derived sentiment vector; and (d) determining whether the specific product review is abnormal based on a difference value between the obtained actual user score and the predicted user score.