چكيده به لاتين
In natural language processing for information extraction, entity linking is a method that has recently attracted the attention of many natural language professionals and researchers. Linking entities performs the task of identifying and disambiguating entities to a knowledge base (such as Wikipedia, DBpedia, or Yago). Binding of entities is also known as NamedEntity Linking(NEL) , NamedEntity Disambiguation(NED) , NamedEntity Recognition and Disambiguation(NERD) , NamedEntity Normalization(NEN) that are responsible to assigning entities to unique links. Enriching text documents with such links removes the ambiguity of the entities. The ambiguities in the texts are a challenge for many text extraction programs, and an entity may be selected by a large number of candidates. In addition, there are many entities that cannot be linked to Wikipedia, because Wikipedia has limited coverage. In this thesis, we create annotated corpus for Arabic entity linking and we examined two ways for entity linking one of them is based and in another one using deep learning algorithms. In first method F¬score was 0.79 and in the second method the accuracy rate was 0.75, which is comparable to the methods available at the English entity linking.
