Recommendation algorithm based on user comments

A technology of user reviews and recommendation algorithms, applied in the field of recommendation algorithms based on user reviews, can solve the problems of poor recommendation effect, lack of new user purchase records, and user recommendation, etc., and achieve the effect of improving the effect.

CN110648163APending Publication Date: 2020-01-03SUN YAT SEN UNIV
3 Cites 18 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2020-01-03

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
Patent Text Reader

Abstract

The invention provides a recommendation algorithm based on user comments. The method is based on a comment recommendation system of deep learning. according to past browsing purchase and comment records of a user, score records of commodities and comments of customers who have purchased the commodities, the preference degree of the user for the commodities is speculated, finally, the recommendation system ranks scores, and the first N commodities with high scores are recommended to the user. According to the method, the comment text information is vectorized by utilizing a natural language processing technology; the importance of each comment and the influence of each comment on the potential customer are calculated according to an attention mechanism algorithm; a convolutional neural network is adopted to capture an interaction feature matrix of a user commodity, so that an interaction vector of the user commodity is generated, finally, the generated interaction vector is input into an FM prediction machine to predict a score, the higher the prediction score is, the more likely the user likes the commodity, and finally, sorting is performed according to the score to recommend N commodities before scoring to the user.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The present invention relates to the technical field of recommendation algorithms, and more specifically, to a recommendation algorithm based on user comments. Background technique

[0002] Recommendation services are ubiquitous in people's daily life, and they can help users discover potentially interesting products or useful services. The recommendation system first collects the user's historical behavior data, then generates the user product rating matrix, and then uses the relevant recommendation technology in the field of machine learning to form a personalized recommendation for the user. A good recommendation system can not only help users find what they like, but also increase the sales and transaction conversion rate of shopping websites, and increase customer loyalty to the website. In the past few years, traditional recommender systems have focused on user-item interactions, such as ratings, purchases, or click records. For example, the ma...

Examples

Embodiment 2

[0052] Such as figure 2 As shown, the present invention provides a recommendation algorithm based on user comments:

[0053] First of all, the recommendation system is to help users find the products they may like, so we need our users' shopping records and product records, all of which use Yelp's public data set, which contains more than 3 million interactive data, each The interaction data includes user ID, item ID, rating and review information. The rating of the item is between 1 and 5. The higher the score, the better the user's evaluation of the item. The review-based recommendation system recommends products that users may like based on the user's shopping history and review information. Therefore, it is necessary to predict the user's rating for each product and recommend the top N products with high ratings to the user.

[0054] The specific method steps are as follows

[0055] 1) Preprocess the Yelp data set, keep only words with occurrences higher than 5, and fil...