SU-NLP at SemEval-2022 Task 11: Complex Named Entity Recognition with Entity Linking

TL;DR

This paper presents an unsupervised entity linking system for complex named entity recognition that leverages Wikipedia context, significantly improving performance in low-context scenarios.

Contribution

The authors introduce a novel unsupervised entity linking pipeline utilizing Wikipedia to enhance complex NER, especially in low-context environments.

Findings

Significant performance improvement in complex entity recognition

Effective use of Wikipedia context for entity classification

Enhanced detection in low-context settings

Abstract

This paper describes the system proposed by Sabanc{\i} University Natural Language Processing Group in the SemEval-2022 MultiCoNER task. We developed an unsupervised entity linking pipeline that detects potential entity mentions with the help of Wikipedia and also uses the corresponding Wikipedia context to help the classifier in finding the named entity type of that mention. Our results showed that our pipeline improved performance significantly, especially for complex entities in low-context settings.