An Approach for Named Entity Disambiguation with Knowledge Graph
Entity disambiguation has always been a key issue in the field of semantic analysis, question answering and recommendation system. The existing approaches for entity disambiguation are based on similarity calculation. The similarity is calculated by considering the similarity of entity context, or t...
Saved in:
Published in | 2018 International Conference on Audio, Language and Image Processing (ICALIP) pp. 138 - 143 |
---|---|
Main Authors | , , , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
01.07.2018
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Entity disambiguation has always been a key issue in the field of semantic analysis, question answering and recommendation system. The existing approaches for entity disambiguation are based on similarity calculation. The similarity is calculated by considering the similarity of entity context, or the correlation between entities. These similarity calculation approaches are used to calculate the entity similarity at paragraph and document level. If the similarity of entities in short text with limited context is needed, the existing methods are not applicable. Therefore, we proposed a disambiguation method based on semantic similarity of ambiguous word. Firstly, according to the entities in the context of ambiguous word, a classifier is constructed to predict the classification of the ambiguous word, and a list of candidate entities is obtained according to the classification. Then, the contextual Resource Description Framework(RDF) triples related to ambiguous words in Knowledge Graph are mapped to the same vector space with the RDF triples related to the entities in the candidate entity table. Finally, semantic similarity is obtained according to the cosine similarity, and the top-k similarity is selected. In this paper, the validity of this method for entity disambiguation is evaluated by data sets. The results show that the proposed method is superior to the existing context similarity baseline. In addition, the improved method is suitable for the most of Knowledge Graph. |
---|---|
DOI: | 10.1109/ICALIP.2018.8455418 |