Environmental and life sciences observations in knowledge graphs using NLP techniques to support multidisciplinary studies

2022

Article

Arslan M., Desconnets Jean-Christophe, Mougenot I.

Procedia Computer Science, 2022, 201, 543-550 ISSN 1877-0509

International Conference on Emerging Data and Industry 4.0. (EDI40), 5., Porto (PORT), 2022/03/22-25

The understanding of environmental observations is a continuous challenge for environmental and life science investigations. The environmental data is complex as it involves its own features, methods, properties, systems, and spatio-temporal dimensions. The time granularity remains approximately the same for different environmental contexts but geographic and rest of the above-mentioned entities are defined using domain vocabularies that are specific for each discipline. It is time-consuming for the researchers of life sciences' discipline to discover, access, and analyze relevant environmental observations as each discipline has its data formats, vocabularies, and metadata standards. These differences introduce structural and semantic heterogeneities, resulting in creating a barrier for reusing datasets generated by other disciplines. Existing dataset discovery platforms contain domain-specific metadata descriptions for explaining datasets which limits their usage. To overcome this knowledge barrier, this work reports the proof-of-concept implementation of a knowledge graph that is centered towards the oceanography use case scenario using NLP techniques (named entity recognition (NER) followed by text preprocessing). The constructed knowledge graph is a collection of subgraphs each representing the metadata of a dataset. It uses the geo-spatial and open semantic data standards that aim to provide enhanced metadata descriptions of datasets for enabling multidisciplinary research.

Informatique [122]

Fonds IRD [F B010092365]

fdi:010092365