An Experiment in Annotating Animal Species Names from ISTEX Resources

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4509–4513 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4509 An Experiment in Annotating Animal Species Names from ISTEX Resources Sabine Barreaux, Dominique Besagni INIST – CNRS 2, rue Jean Zay 54519 Vandœuvre-lès-Nancy {sabine.barreaux, dominique.besagni}@inist.fr Abstract To exploit scientific publications from global research for TDM purposes, the ISTEX platform enriched its data with value-added information to ease access to its full-text documents. We built an experiment to explore new enrichment possibilities in documents focussing on scientific named entities recognition which could be integrated into ISTEX resources. This led to testing two detection tools for animal species names in a corpus of 100 documents in zoology. This makes it possible to provide the French scientific community with an annotated reference corpus available for use to measure these tools’ performance. Keywords: named entity recognition, zoology, ISTEX, Text and Data Mining 1. Introduction The availability of an ever-increasing volume of scientific publications requires a number of means to automatically mine into large sets of documents and extract implicit knowledge and correlations to be used in research work. But before being able to carry out this text search, one needs to address the questions of accessing these documents in full text and of pre-processing them, the quality of the documents being often uneven. The ISTEX platform (Initiative d'Excellence en Information Scientifique et Technique)1 offers a solution to the French higher education and research community by providing access to retrospective collections of scientific literature in all disciplines (Colcanap, 2013). This vast multidisciplinary and multilingual collection contains more than 23 million scientific publications and is constantly enriched with information to improve the text quality to optimize machine processing, i.e. categorization of documents by scientific fields, extraction and structuring of bibliographic references, detection of named entities, structuring of the full text in XML-TEI from PDF (Collignon and Cuxac, 2017). This platform provides massive value-added downloads and represents an unparalleled resource for text mining research (Cuxac and Thouvenin, 2017). In this article, we will describe an experiment led to continue the work on ISTEX data enriching by testing two tools for the detection of a new type of named entities, animal species names. We will trace the steps of building and annotating a reference corpus in zoology from ISTEX resources and compare the performance of the two tools tested: entity-fishing2 and IRC3sp. We will conclude by considering the possibilities of using these tools in ISTEX. 2. Use Cases Users can run queries in the ISTEX platform on traditional bibliographic fields, but also on fields that exploit the value-added information injected into documents. Among the enhancements to ISTEX documents, we will focus on named entities. This enrichment was provided by 1 Excellence initiative in scientific and technical information: https:www.istex.fr 2 http://nerd.huma-num.fr/nerd/ implementing the UNITEX CasSys tool at INIST3 in collaboration with LIFAT4. At present, this tool has made it possible to detect the named entities contained in nearly 16 million documents …

Content truncated to 500 words