Proto-OKN theme 1: knowledge graph to support evaluation and development of climate models

This Cooperative Agreement award from the National Science Foundation (NSF) under the Technology, Innovation, and Partnerships (TIP) program (CFDA 47.084) provides $1,498,564 to Temple University to develop a large, multimodal knowledge graph that captures the salient aspects of climate modeling, including data, models, and analytical tasks. The 3-year project aims to create automated methods for inferring relationships between research papers, datasets, climate models, and tools, and incorporate these insights into an intuitive, structured knowledge graph. This will facilitate the discovery and reuse of existing climate research artifacts, as well as foster collaboration and innovation in the field. The award utilizes novel deep learning techniques for entity and relation extraction, entity linking, and knowledge graph construction. The project also has a focus on training and educating students at various academic levels. No sub-awards are planned under this grant.

Abstract

This prototype-open knowledge network project on knowledge graph to support evaluation and development of climate models, aims to create a large multimodal knowledge graph of the most salient aspects of climate modeling, including data, climate models, and tasks. The climate models covered include both classical fluid dynamics models as well as AI-based models. The knowledge graph that is developed will take a holistic view of climate modeling while addressing one of the three key problems identified in the report of the office of science roundtable on data for AI, VIZ., addressing ?open questions in AI with frameworks for relating data, models, and tasks.? The climate model knowledge graph will ensure that existing models and datasets are leveraged in new climate modeling undertakings, thereby ensuring that past research investments are reused and fully leveraged in future work. The automated methods developed in this project will help infer paper-data-model-tasks relations, providing the ability to suggest useful, related artifacts to new undertakings thus shortening the time for relevant artifact searches. Successful creation of paper-model-tool relations and embedding of those into a knowledge graph will help provide a structured representation of climate models, making them more easily accessible. The project contributes significantly to the field of information retrieval, with a particular focus on named entity recognition and the creation of a comprehensive knowledge graph. Research papers provide the necessary context for reusing research artifacts, by linking together data, models, and analyses, describing the role of a dataset (e.g., training or testing) and indicating whether a model is original or used as a baseline. Incorporating these insights into a knowledge graph provides researchers an intuitive and structured means for navigating the complex relationships among datasets, models, tools, and methods. This not only facilitates the discovery and reuse of existing research artifacts but also fosters collaboration and innovation in the field research, in this case climate research. Novel deep learning techniques will be developed for automatic entity and relation extraction, entity linking, and construction of a knowledge graph interconnecting these aspects. Novel technology will be developed for automatically identifying and cataloging public climate data and related highly reusable tools. The proposed approach is multimodal, dealing with images as well as tables from text-based publications. This project also has a focus on teaching and training of students at various levels, from high school to doctoral programs. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.