Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Conversational recommendation method and system based on graph neural network and review similarity

A recommendation method and neural network technology, applied in the fields of artificial intelligence and recommendation systems, can solve problems such as the inability to accurately capture the complex dependencies of commodities in the session, the inability to accurately model user preferences, and the difficulty in obtaining recommendation effects, etc., to increase reliability Effects of interpretability, optimized weights, and improved accuracy

Active Publication Date: 2022-07-05
南方电网互联网服务有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the method based on graph neural network is only based on the co-occurrence frequency or transition probability of commodities when defining the commodity relationship. With the help of the flexibility of the graph structure, it alleviates the strict time sequence dependence of the recurrent neural network session recommendation; however, its Does not take into account that in real shopping scenarios, a session may contain multiple user intents
This makes existing methods unable to accurately capture the complex dependencies between items in a session
To sum up, the existing methods cannot accurately model the relationship between products and user preferences, and it is difficult to achieve satisfactory recommendation results.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Conversational recommendation method and system based on graph neural network and review similarity
  • Conversational recommendation method and system based on graph neural network and review similarity
  • Conversational recommendation method and system based on graph neural network and review similarity

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0103] The overall model framework of the present invention is as follows figure 2 shown. Depend on figure 2 It can be seen that the main frame structure of the present invention includes a global graph module, a local graph module, a session generation module and a candidate product prediction module based on the similarity of reviews. Among them, the global graph module based on comment similarity first uses the idea of ​​graph attention network to generate attention weights according to the importance of each connection; the global graph based on comment similarity is the global graph neighbor information obtained after preprocessing and adjacent edge weights; after the global graph is constructed, the nodes in the graph, that is, the commodities in the session, are updated and learned by means of a graph attention neural network, and the final weighted summation is used to obtain the global graph commodity representation. The local graph module uses the paired commodit...

Embodiment 2

[0105] The specific steps of the conversation recommendation method based on graph neural network and comment similarity of the present invention are as follows:

[0106] S1. Session sequences required for establishing a training set and a test set: First, it is necessary to obtain user interaction records on an e-commerce website, and then perform preprocessing operations on them to obtain session sequences that meet the training requirements. The specific steps are as follows:

[0107] S101. Download a data set of an e-commerce website that has been published on the Internet, and use it as raw data for constructing a session sequence.

[0108] For example: There are many publicly available historical datasets of user behavior for recommender systems, such as Amazon's Pet Supplies dataset.

[0109] The data format in the Pet Supplies dataset is as follows:

[0110]

[0111] Where reviewerID is the user ID; asin is the product ID; reviewerName is the user's nickname; review...

Embodiment 3

[0240] Based on the conversational recommendation system based on graph neural network and comment similarity according to Embodiment 2, the system includes:

[0241] The session sequence construction unit of the training set and the test set first needs to obtain the user's interaction records on the e-commerce website, and then preprocess them to obtain the session sequence that meets the training requirements; the session sequence construction unit of the training set and the test set include,

[0242] The original data acquisition unit is responsible for downloading the data sets of e-commerce websites that have been published on the Internet as the original data for constructing the session sequence;

[0243] The original data preprocessing unit is responsible for setting the time span of the session to a certain time period, and constructing the training set and test set session sequence that meets the conditions, so as to construct the session sequence of the training s...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a session recommendation method and system based on graph neural network and comment similarity, and recommends target commodities for users in the current session according to commodity sequence and comment information in the session. The invention integrates the comment information into the graph neural network model, and simultaneously considers the inter-commodity dependency relationship contained in the conversation and the inter-commodity similarity relationship in the text space. This conversational recommendation method is mainly composed of four modules: the global graph module based on the similarity of the reviews, which obtains the similarity of the product in the text space according to the review document of the product, and constructs the global map of the product based on the reviews according to the similarity; the local graph module , according to the product sequence of the current session, obtain the local map of the product in the session; the session generation module generates the final session representation by combining the product global map and the product local map representation obtained by the product in the first two modules respectively; the candidate product prediction module, According to the session representation, the score of each candidate item is predicted, and the target item is recommended.

Description

technical field [0001] The present invention relates to the technical fields of artificial intelligence and recommendation systems, in particular to a conversation recommendation method and system based on graph neural network and comment similarity. Background technique [0002] With the rapid development of e-commerce, the number of commodities has increased rapidly, and the original search engine method is difficult to help users find the most desired target commodity among the massive candidate commodities, resulting in the problem of information overload. Recommender system is an important method to solve this problem, it can accurately find the target product in a large number of candidate products and recommend it to the user. At present, many online business platforms have weakened the search function and mainly rely on the recommendation function, which greatly reduces the user threshold, such as Taobao and JD.com. Recommender systems can not only provide more conv...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06Q30/06G06F16/335G06F16/9535G06N3/04G06N3/08
CPCG06Q30/0631G06F16/9535G06F16/335G06N3/08G06N3/045Y02D10/00
Inventor 鹿文鹏张骞邵珠峰王荣耀
Owner 南方电网互联网服务有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com