Research on Chinese Medicinal Materials News Recommendation Method Based on Knowledge Graph
As the"Internet+"era unfolds,numerous Chinese medicinal materials information platforms and mobile applications promptly deliver industry-related news to both sellers and buyers in this field.In contrast to conventional news recommendations,the context of Chinese medicinal materials news recommendation demonstrates a notable focus on user specificity.Moreover,the depth of domain knowledge is more profound,with nearly every news headline intricately linked to information about Chinese medicinal materials.Consequently,the integration of Chinese medicinal materials knowledge into the news recommendation process is pivotal for elevating the efficacy of these recommendations.However,in traditional Chinese medicine(TCM)classics,"knowledge"is stored in a semi-structured form,akin to dictionaries.This storage method poses a challenge for retrieval,dissemination,and application of knowledge.The application of knowledge graph technology to the field of Chinese medicinal materials can address this issue by enabling structured storage and visual representation of knowledge.This,in turn,establishes a knowledge foundation for Chinese medicinal materials news recommendation.Additionally,conventional news recommendation methods often focus solely on the interaction between users and news,along with the textual information of the news.Applying these methods directly to the specific domain of Chinese medicinal materials news recommendation will neglect the potential connections between Chinese medicinal knowledge and the aforementioned information.To address the issues mentioned earlier and enhance the client's ability on the Chinese medicinal materials information platforms and mobile applications to offer users precise news recommendations and an improved user experience,this paper presents a method of Chinese medicinal materials news recommendation based on knowledge graph.First,entities and relationships are extracted from the classics of Chinese medicinal materials to construct the knowledge graph of Chinese medicinal materials.Second,the news headlines are segmented to linked to the relevant knowledge graph entities.On this basis,combined with word embeddings and entity embeddings,the convoluted news embeddings are obtained through the Attention-based Knowledge-aware Convolutional Neural Networks(AKCNN).Third,based on the user's historical click news and side information,the Top-K recommendation list is generated by using the Chinese medicinal materials news recommendation model based on knowledge graph and attention network.To demonstrate the effectiveness of the proposed method in this paper,we utilize widely accepted metrics such as HR@10 and MAP@10 from the news recommendation field as evaluation criteria.We conduct a comparative analysis between the proposed method and five baseline models,using two real datasets.Experimental results show that the HR@10(0.750,0.788)and MAP@10(0.468,0.471)of the proposed method are better than baseline methods.This method imports the knowledge graph into Chinese medicinal materials news recommendation,which effectively improves the quality of recommendation and has certain interpretability.In theoretically terms,this paper innovates by proposing entity and relationship extraction methods tailored to different types of entities and relationships based on the characteristics of TCM materials literature.This enhances the structured storage of TCM materials knowledge,introducing novel approaches for more convenient storage and utilization.In practical terms,a knowledge graph-based method for news recommendation in the field of TCM materials is introduced,contributing to the effectiveness of vertical domain news recommendation algorithms.This method not only complements existing research methodologies but also offers more precise and compelling news recommendations for users,including herbal farmers and pharmaceutical merchants.Moreover,it provides a valuable reference for integrating TCM materials knowledge into the design of recommendation algorithms for platforms focusing on TCM materials information.
NETL
NSTL
万方数据
