Effective Personalized Taxonomy Pruning for Identification of Relevant Knowledge Domain for Knowledge Support of User Interests

Although living in the information era, users are still more likely to have partial information rather than comprehensive understanding on some concepts due to limited knowledge. In order to provide user with knowledge support of his interests, we propose an efficient personalized pruning model called TP-Alpha for the pruning of taxonomy graph while quantifying the relation between connected entities based on introduced concept called as relevancy degree. TP-Alpha model is a variant of approximation algorithm for acyclic directed Steiner tree problem. Compared to state-of-art pruning models, TP-Alpha model can preserve most of the user-given interests and entities with relevant knowledge having maximum average of relevancy degree inferred by user-given threshold value, and the least of other entities for maintaining connectivity without the constraint of the location of user-interested entities in the taxonomy graph. We evaluate our model on the benchmark datasets from a SemEval challenge. Experimental results verify the effectiveness of our TP-Alpha model in achieving relevant knowledge domain with more comprehensive knowledge support of user interests while maintaining less unnecessary cost.