The Research On Coreference Resolution Based On Knowledge Base And Entity Linking

The era of information explosion had come. Information extraction became particularly important in the mass data. Entity linking is the process to link the mention with context to the node of knowledge base. Coreference resolution is the process to judge whether the different mention refer to the same entity. These techniques were important to intent recognition of query, information extraction and construction of knowledge base automatically.This paper studied entity linking and coreference resolution, the main content is as follows,1. This paper proposed and implemented a model for entity linking which is based on learning to rank. This method used CRF to extraction mention. Sort the candidate nodes based on learning to rank.2. This paper proposed and implemented the word of force model to calculate the similarity between nodes of KB. We proposed and implemented a random walk model to calculate the weight of KB node. We used these features to optimize the result.3. This paper proposed and implemented an entity linking algorithm which is based on coreference resolution. This method used CRF to extraction mention, implemented the system of coreference resolution, employed the word of force to optimize the entity linking.This paper designed comparative experiment on Chinese micro-blog entity linking corpus and English entity linking corpus. The accuracy rate is about80.7%on the corpus of Chinese micro-blog entity linking, has an increase of8percents comparing with baseline.