Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised Named Entity Disambiguation for Low Resource Domains

Debarghya Datta, Soumajit Pramanik

TL;DR

This paper introduces GST-NED, an unsupervised approach for named entity disambiguation in low-resource, domain-specific domains. By constructing candidate graphs and solving minimum-cost Group Steiner Trees over candidate entities, it jointly leverages cross-mention context to identify gold entities without annotated data. The method demonstrates substantial improvements (over $40\%$ on average in Precision@1) over unsupervised baselines across diverse domains, with GST-count as a robust ranking signal. Limitations include dependence on sufficient entity density and scalability considerations for larger knowledge graphs, while future work points to improved scalability and handling longer documents. The approach holds practical value for humanities, technical writing, and biomedical contexts where domain KBs are small and labeled data are scarce.

Abstract

In the ever-evolving landscape of natural language processing and information retrieval, the need for robust and domain-specific entity linking algorithms has become increasingly apparent. It is crucial in a considerable number of fields such as humanities, technical writing and biomedical sciences to enrich texts with semantics and discover more knowledge. The use of Named Entity Disambiguation (NED) in such domains requires handling noisy texts, low resource settings and domain-specific KBs. Existing approaches are mostly inappropriate for such scenarios, as they either depend on training data or are not flexible enough to work with domain-specific KBs. Thus in this work, we present an unsupervised approach leveraging the concept of Group Steiner Trees (GST), which can identify the most relevant candidates for entity disambiguation using the contextual similarities across candidate entities for all the mentions present in a document. We outperform the state-of-the-art unsupervised methods by more than 40\% (in avg.) in terms of Precision@1 across various domain-specific datasets.

Unsupervised Named Entity Disambiguation for Low Resource Domains

TL;DR

This paper introduces GST-NED, an unsupervised approach for named entity disambiguation in low-resource, domain-specific domains. By constructing candidate graphs and solving minimum-cost Group Steiner Trees over candidate entities, it jointly leverages cross-mention context to identify gold entities without annotated data. The method demonstrates substantial improvements (over on average in Precision@1) over unsupervised baselines across diverse domains, with GST-count as a robust ranking signal. Limitations include dependence on sufficient entity density and scalability considerations for larger knowledge graphs, while future work points to improved scalability and handling longer documents. The approach holds practical value for humanities, technical writing, and biomedical contexts where domain KBs are small and labeled data are scarce.

Abstract

In the ever-evolving landscape of natural language processing and information retrieval, the need for robust and domain-specific entity linking algorithms has become increasingly apparent. It is crucial in a considerable number of fields such as humanities, technical writing and biomedical sciences to enrich texts with semantics and discover more knowledge. The use of Named Entity Disambiguation (NED) in such domains requires handling noisy texts, low resource settings and domain-specific KBs. Existing approaches are mostly inappropriate for such scenarios, as they either depend on training data or are not flexible enough to work with domain-specific KBs. Thus in this work, we present an unsupervised approach leveraging the concept of Group Steiner Trees (GST), which can identify the most relevant candidates for entity disambiguation using the contextual similarities across candidate entities for all the mentions present in a document. We outperform the state-of-the-art unsupervised methods by more than 40\% (in avg.) in terms of Precision@1 across various domain-specific datasets.

Paper Structure

This paper contains 11 sections, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Problem Statement
  3. Methodology
  4. Candidate Generation
  5. Candidate Ranking
  6. Experimental Setup
  7. Datasets
  8. Baselines
  9. Metrics
  10. Results and Discussion
  11. Conclusion

Figures (1)

  • Figure 1: Proposed GST-NED approach: The sample document at the top contains three mentions; the subgraph extracted from the KB is shown at the left and the minimum cost GST is shown using the box at the right. In the induced subgraph, the candidates for 'adenoma' are marked in red, 'hamartomas' are marked in yellow and 'liver rupture' are marked in blue.