Method and apparatus for performing entity linking

An apparatus used for performing entity linking between a surface entity mention in a surface text and entities of a knowledge graph, the apparatus comprising:

• a computer-implemented contextual text representation model adapted to calculate a contextual representation for each token within a received surface text and having a pooling unit adapted to pool contextual representations of the tokens of a surface entity mention in the surface text with contextual representations of other tokens within the surface text to provide a contextual entity representation vector representing the surface entity mention with a surface text context;
• a computer-implemented knowledge graph embedding model, KGEM, adapted to provide an entity node embedding vector for a received candidate entity identifier of a candidate knowledge graph entity, selected for the surface entity mention using a method comprising a sub-linear query cost as a result of a candidate selection process forming one of a group of top-matched candidate entity node identifiers, for the surface entity mention;
• and
• a processing unit adapted to combine the contextual entity representation vector with the entity node embedding vector to generate an input vector applied to a fully connected layer which provides an unnormalized output transformed by a softmax function into a normalized output processed to classify whether the surface entity mention corresponds to the candidate knowledge graph entity and to output a corresponding label.