Named Entity Linking with Entity Representation by Multiple Embeddings

TL;DR

This paper introduces a simple method for named entity linking using multiple embeddings for entity representation, demonstrating its effectiveness on challenging datasets and analyzing the impact of various parameters.

Contribution

The paper presents a practical approach to NEL with multiple embeddings, showing that a small number of embeddings suffices and that tuning on diverse news improves performance.

Findings

Minimal mentions needed for KB entities significantly affect NEL performance.

Using as few as 10 embeddings can be effective.

Tuning embeddings on diverse news yields better results.

Abstract

We propose a simple and practical method for named entity linking (NEL), based on entity representation by multiple embeddings. To explore this method, and to review its dependency on parameters, we measure its performance on Namesakes, a highly challenging dataset of ambiguously named entities. Our observations suggest that the minimal number of mentions required to create a knowledge base (KB) entity is very important for NEL performance. The number of embeddings is less important and can be kept small, within as few as 10 or less. We show that our representations of KB entities can be adjusted using only KB data, and the adjustment can improve NEL performance. We also compare NEL performance of embeddings obtained from tuning language model on diverse news texts as opposed to tuning on more uniform texts from public datasets XSum, CNN / Daily Mail. We found that tuning on diverse news provides better embeddings.