Construction d'un système de recommandation basé sur des contraintes via des graphes de connaissances

Ngoc Luyen Le,Marie-Hélène Abel,Philippe Gouspillou

2023-06-06

Abstract:Knowledge graphs in RDF model entities and their relations using ontologies, and have gained popularity for information modeling. In recommender systems, knowledge graphs help represent more links and relationships between users and items. Constraint-based recommender systems leverage deep recommendation knowledge to identify relevant suggestions. When combined with knowledge graphs, they offer benefits in constraint sets. This paper explores a constraint-based recommender system using RDF knowledge graphs for the vehicle purchase/sale domain. Our experiments demonstrate that the proposed approach efficiently identifies recommendations based on user preferences.

Information Retrieval

What problem does this paper attempt to address?

This paper attempts to address the problem of how to provide more accurate product recommendations based on user preferences and needs in high-value, low-frequency purchase scenarios (such as car buying and selling). Specifically, the authors propose a recommendation system based on constraints and implemented through a knowledge graph, aiming to improve the accuracy and relevance of recommendations by combining domain-specific knowledge (such as vehicle attributes) and user personal preferences (such as color, number of seats, etc.).

In traditional recommendation systems, such as those based on collaborative filtering, content-based methods, or demographic-based methods, it may be challenging to provide high-quality recommendations when data is limited, especially in high-value product domains where users need to explicitly express their needs. Therefore, this study focuses on constructing a recommendation system capable of handling specific user requirements and utilizing domain knowledge for reasoning to overcome issues like cold start and data sparsity in traditional methods.

The recommendation system described in the paper addresses the problem mainly through the following ways:

1. **Construction of the Knowledge Graph**: Using RDF (Resource Description Framework) to build a semantic network containing vehicle attributes, user information, and their interactions, forming a structured knowledge base.
2. **Constraint-based Recommendation Mechanism**: Defining two types of constraints—domain knowledge constraints and user preference constraints—and applying these constraints to the knowledge graph through logical rules (SWRL) and query language (SPARQL) to identify recommendations that meet user needs.
3. **Experimental Validation**: Conducting experiments on an RDF knowledge graph containing a large number of vehicle descriptions and user preferences to validate the effectiveness and accuracy of the proposed recommendation system.

In summary, this paper explores an innovative design approach for recommendation systems, particularly in handling high-value product recommendations, by effectively integrating domain knowledge and user personalized needs to provide more accurate services.