This repository contains the code for our paper titled "Unsupervised Named Entity Disambiguation for Low Resource Domains", accepted at EMNLP 2024.
The methodology introduces an unsupervised approach for Named Entity Disambiguation (NED) using fuzzy text search and graph-based candidate selection techniques. This approach is tailored for low-resource settings where labeled data is scarce or unavailable.
The code for the NED pipeline will be updated soon. Stay tuned!
In the ever-evolving landscape of natural language processing and information retrieval, the need for robust and domain-specific entity link- ing 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 re- source 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 candidate 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 40-50% in terms of Precision@1 across various domain-specific datasets.
Instructions for installing the required dependencies will be provided in the upcoming release.
Detailed usage examples will be available soon. Check back later for more updates.
For any questions or feedback, feel free to open an issue or reach out via email at debarghyad@iitbhilai.ac.in.