Computing semantic relatedness for domain entities from encyclopaedias of digital publishing resource

摘要

The task of computing semantic relatedness for domain-specific entities may be seen as one of the pivotal processes for implementation of the domain-specific knowledge service. Encyclopedias of digital publishing resources are a kind of reference book which generally holds a relatively complete and authoritative collection of entities in a specific domain. In this paper, a novel method named Globally Explicit and Implicit Analysis (GEIA) is proposed to compute semantic relatedness for entities created from encyclopedias. First, a weighted directed graph is built with the entities as vertexes and explicit relations between the entities as edges. The implicit relations between the entities are then computed from the graph based on the explicit relations extracted from the encyclopedias. With this method, both the explicit and implicit relations are used to compute the semantic relatedness. Experiment results show that the method is effective.