Music Recommendation Method Study Based On Ontology Modeling And Context Awareness

With the popularity of information technology,It becomes more and more important to quickly and accurately acquire the useful data from the massive data.Personalized recommendation is a technique that helps users to efficiently find the interested information in the case of information overload.The current "ubiquitous network" state makes the context-aware recommendation system(CARS)become a research hotspot.The context-aware system can use contextual information to further improve the recommended accuracy.In recent years,streaming media services are becoming more and more popular and people usually choose their favorite playlist to listen to music.Therefore,how to recommend the appropriate playlist to the user is already the center of the music recommendation.User preference modeling,music resources and recommendation algorithms constitute the three basic elements of music recommendation.At present,the research of music recommendation mainly improves the recommendation accuracy by optimizing the recommendation algorithm,ignoring the particularity of the music resource itself and the important connection between the user's song and the situation information.To this end,this paper proposes a music recommendation method based on ontology modeling and context perception to explore the use of music resources to better characterize the user's preferences,and to recommend some special music according to the user's taste.The main work of this paper includes:(1)Build a music-friendly service-based ontology(CHMO)to process music knowledge which is compatible with Chinese songs and listeners.In view of the increasing demands of music service quality by the Internet especially the mobile Internet,the lack of knowledge in the field of music and the differences between Chinese and Western music aesthetic,this paper firstly designs a fine-grained music ontology model,using Protégé and OWL to build music ontology based on reasonable construction principles and methods;Then,the composition and model of the ontology are elaborated,and the reasoning model is optimized by Jena reasoning machine.Finally,we focus on the correct detection of the algorithm in ontology and the supportive problem in ontology implementation.(2)The algorithm of multi-layer collaborative filtering music recommendation based on ontology modeling and tensor decomposition is designed and implemented.In order to try to solve the semantic gap between low-level audio features and deep-seated music comprehension,and the influence of situational factors,In this paper,we use the ontology to construct the user knowledge model,combine the multi-case similarity measure,through the preliminary screening of candidate neighbor sets,Then we construct the time-based weight tensor score based on the user's activity,use the tensor decomposition to get the recommended evaluation value,and finally use the recommended evaluation value for Top-N music recommendation.The experimental results show that the algorithm can get a better recommendation in the case of extremely sparse data.