Label Distribution Learning-Based Semantic Retrieval Model on Knowledge Graph

The traditional retrieval technology is difficult to grip users’ intention accurately and apace in massive and fragmented data. To organize fragmented knowledge in reason and retrieve useful knowledge to recommend for users in a big data environment, this paper proposes a label distribution learning-based semantic retrieval model on the knowledge graph. The model firstly constructs the knowledge graph based on the topic map model. And then it clusters knowledge by using K-means label distribution learning (KM-LDL) and reduces retrieval scope. To address the problem that the space vector converted from the input text is sparse, this paper improves K-means label distribution learning according to the idea of probability and constructs a semantic retrieval model on a knowledge graph based on Bayes-K-means label distribution learning (Bayes-KM-LDL). The experimental results show that the proposed model can organize fragmented knowledge more reasonably than the traditional retrieval approach and its accuracy and recall rate have good performance.