EsPharmaNER

This repository contains the source code and data to reproduce the results of Stoeckel et. al (2019) on the PharmaCoNER Challenge.

Setup

• Extract the CoNLL 2003 formatted datasets from /resources/data.7z into /resources/.
• Download the test data from here and extract the contents of the test-set_1.1.zip folder to gold/.
  • The relative path for gold annotated data should now be gold/test/subtrack1/.

Hardware Pre-requisites

You will need a CUDA compatible GPU with ~6GB VRAM available. During our experiments we used a NVIDIA GeForce GTX 1660.

Dataset

PharmaCoNER Corpus

The PharmaCoNER dataset was released as part of the PharmaCoNER Challenge. It contains 1000 clinical cases from spanish open access literature with gold standard annotations for four categories. For more information, refer to the challenge organizers website: http://temu.bsc.es/pharmaconer/index.php/datasets/

Spanish Health Corpus

The Spanish Health Corpus is comprised of a selection of Spanish Health Science documents. These documents were obtained from SciElo by means of an automated crawler. The full list of document ids that were used to create the corpus can be found in the Spanish_Health_Corpus-document_ids.txt file that is part of this repository.

These IDs can be used to obtain the document text from its corresponding collection like this: http://scielo.isciii.es/scielo.php?script=sci_arttext&pid=S0004-05922010000100016

Please note, that the IDs must be mapped to the correct collection to obtain the documents. See https://scielo.org/ or the list below for a all available collections:

• http://www.scielo.org.ar
• http://www.scielo.org.bo
• http://www.scielo.br
• http://www.scielo.cl
• http://www.scielo.org.co
• http://www.scielo.sa.cr
• http://scielo.sld.cu
• http://www.scielo.org.mx
• http://scielo.iics.una.py
• http://www.scielo.org.pe
• http://www.scielo.mec.pt
• http://www.scielosp.org
• http://www.scielo.org.za
• http://scielo.isciii.es
• http://www.scielo.edu.uy
• http://scielo.senescyt.gob.ec
• http://ve.scielo.org
• http://westindies.scielo.org

Cite

Please use the following citation:

M. Stoeckel, W. Hemati, and A. Mehler, "When Specialization Helps: Using Pooled Contextualized Embeddings to Detect Chemical and Biomedical Entities in Spanish", in Proceedings of the International Workshop on BioNLP Open Shared Tasks (BioNLP-OST), 2019. accepted

BibTex entry:

@InProceedings{Stoeckel:Hemati:Mehler:2019,
    author = {Manuel Stoeckel and Wahed Hemati and Alexander Mehler},
    title = {{When Specialization Helps: Using Pooled Contextualized Embeddings to Detect Chemical and Biomedical Entities in Spanish}},
    booktitle = {Proceedings of the International Workshop on BioNLP Open Shared Tasks (BioNLP-OST)},
    publisher = {Association for Computational Linguistics SIGDAT and Asian Federation of Natural Language Processing},
    location = {Hong Kong, China},
    year = 2019,
    note = {accepted}
}

Acknowledgements

This publication is part of the BIOfid project. Visit https://www.biofid.de/en/ for more information.