IN030-10
Es2Vec: Earth Science Metadata Suggestions and Analogical Reasoning
Monday, 14 December 2020: 04:27
Virtual
Carson Davis1, Muthukumaran Ramasubramanian2, Derek Koehl3, Ashish Acharya2, Iksha Gurung2, Manil Maskey4, Rahul Ramachandran5 and Udaysankar S Nair6, (1)Manufacturing Technical Solutions, Huntsville, AL, United States, (2)University of Alabama in Huntsville, Huntsville, AL, United States, (3)UAH, Huntsville, AL, United States, (4)NASA Marshall Space Flight Center, MSFC, Huntsville, AL, United States, (5)NASA Marshall Space Flight Center, Huntsville, AL, United States, (6)University of Alabama in Huntsville Earth System Science Center, Huntsville, AL, United States
Abstract:
As the volume of text-based Earth science research grows, it is increasingly possible to discover latent relationships in the literature. However, traditional methodologies are restricted by limited computational capabilities and intractable problem spaces. Advancements in natural language processing (NLP) have allowed us to use a comprehensive Earth science corpus to create a domain-specific word vector model, Es2Vec, which we have used to surface latent relationships between Earth science concepts, to generate improved keyword tags, and to explore the area of analogical reasoning.
By using Es2Vec with cosine proximity and domain filtering, we have successfully predicted a wide range of relationships, such as synonyms for common phenomena and the instruments most associated with particular authors. Additionally, we have built a tool that uses Es2Vec to recommend keyword tags for dataset abstracts, potentially improving dataset search and discovery. Finally, Es2Vec has outperformed general word embeddings in surfacing complex word-pair relationships present in the corpus, such as property-instrument pairs (eg. thermometer - temperature). Success in identifying these word pairs is key to ultimately predicting analogous casual relationships.
With this presentation we will demonstrate the capabilities of Es2Vec in surfacing domain-specific insights into an Earth science corpus.
