Robust Candidate Generation for Entity Linking on Short Social Media Texts
This addresses the challenge of accurately linking entities in noisy, short social media texts for applications in knowledge base integration, though it is incremental as it builds on existing methods with domain-specific improvements.
The paper tackled the problem of entity linking on short, informal social media texts like Tweets, where existing dense retrieval methods struggle due to issues like informal spelling and limited context. The result was a hybrid solution using long contextual representations from Wikipedia, achieving a recall of 0.93, which shows considerable gains over previous work.
Entity Linking (EL) is the gateway into Knowledge Bases. Recent advances in EL utilize dense retrieval approaches for Candidate Generation, which addresses some of the shortcomings of the Lookup based approach of matching NER mentions against pre-computed dictionaries. In this work, we show that in the domain of Tweets, such methods suffer as users often include informal spelling, limited context, and lack of specificity, among other issues. We investigate these challenges on a large and recent Tweets benchmark for EL, empirically evaluate lookup and dense retrieval approaches, and demonstrate a hybrid solution using long contextual representation from Wikipedia is necessary to achieve considerable gains over previous work, achieving 0.93 recall.