Available under a Creative Commons Attribution Non-Commercial Share Alike 4.0 International Licence
Hyperbolic embeddings have become important in many natural language processing tasks due to their great ability to capture latent hierarchical data and to encode valuable syntactic and semantic information. We study and consider the ability of Poincaré embeddings to get the most similar nodes to a given node when trying to recognize named entities in a set of text documents. In this paper, we propose a classifier model for the NER (Named Entity Recognition) task by implementing Poincaré embeddings and by using the most frequent n-grams and their Part-of-Speech (POS) structures from the training dataset. We found that POS structures and n-grams help to map possible named entities, while using Poincaré embeddings manage to affirm and refine this recognition, improving the recognition of named entities.
Muñoz D., Pérez F., Pinto D. (2020) Poincaré Embeddings in the Task of Named Entity Recognition. In: Martínez-Villaseñor L., Herrera-Alcántara O., Ponce H., Castro-Espinoza F.A. (eds) Advances in Computational Intelligence. MICAI 2020. Lecture Notes in Computer Science, vol 12469. Springer, Cham. DOI: 10.1007/978-3-030-60887-3_17