A method for summarizing user reviews based on review usefulness

Abstract

The present invention provides a method for automatic user comment summarization based on comment usefulness, including the following steps: Step 1: preprocessing; morphological restoration of the comment text; Step 2: comment usefulness prediction; Features of review usefulness, use the extracted features to characterize reviews, and use random forest classification model to predict the usefulness of reviews; Step 3: Sentiment-topic modeling based on binary words; adding to the traditional binary word topic model Sentiment variable, modeling both topic and sentiment for comments; Step 4: Multi-factor topic and comment ranking. The beneficial effects of the present invention are: 1. The method of the present invention can effectively utilize some neglected important review features to assist in the prediction of the usefulness of reviews and the subsequent sorting and summarization tasks; 2. The sorting of topics in the method of the present invention can save developers time.

Description

technical field [0001] The present invention relates to the field of Internet technology, in particular to a method for summarizing user comments based on comment usefulness. Background technique [0002] The quality of the application directly affects the user experience and is related to the developer's profitability. As more and more apps emerge, app developers face the challenge of how to stand out from the competition. User reviews are feedback from users after using the app. These comments are of great value to the application version update plan, such as fixing major bugs, adding important features, etc. [0003] User review mining has been widely studied in both academia and industry, mainly focusing on user review ranking, review classification, predicting user likes / dislikes, and identifying application emergent problems. Most of these domains rely on predefined categories and large-scale human annotations. However, for popular applications, user reviews are in...

AI Technical Summary

Problems solved by technology

However, for popular applications, user comments are increasing by tens of thousands, and it is time-consuming to mark whether these comments are informative, so there is an urgent need for automatic summary of comments

The current challenge of review summarization is mainly that user reviews are short and contain a lot of text noise. Only about 30% of the reviews are informative, and the reviews contain many topics, which are difficult to be predefined.

[0004] Previous work has ignored some important comment features, such as the number of support for a comment by other users.

Images