Type: Bachelor or Master project, with a possible extension to a thesis.
Your interests / skills: Efficient programming in C++, Natural Language Processing
Entity Linking and Coreference Resolution are important steps in the pipeline for downstream NLP tasks like Question Answering or Relation Extraction. Entity Linking on Wikipedia articles is different from Entity Linking on other corpora since intra-Wiki hyperlinks that are manually created by the authors of an article already provide some links "for free". Our existing rule-based system uses these hyperlinks together with entity information extracted from Wikidata such as aliases, popularity, type or gender to link mentions in a given Wikipedia article to Wikidata entities.
Goal: Reduce the run time of our Entity Linking + Coreference Resolution tool while maintaining the quality of the linking results, by efficiently re-implementing the tool in C++.