GENERATIVE AI IN SUPPORT OF GEOSCIENCE EDUCATION RESOURCE DISCOVERY

The Science Education Resource Center (SERC) plays host to thousands of educational resources developed by the Geoscience Education community for their collective shared use. One of the key challenges SERC faces as custodian of these collections is supporting community members in finding the particular resource they need in a timely and efficient way. Generative AI, and in particular large-language models (LLM), hold significant potential for solving some of the most challenging problems in resource discovery. Over the past two years, SERC has developed internal expertise in the best practices around the use of LLMs that are emerging in the larger search industry. We will describe our experiences in applying a range of these approaches. This includes work to develop and deploy hybrid embedding-based/lexical search tools across our existing collections, as well as other experiments in deploying LLM-based tools to support geoscience educators in exploring and understanding our collections. As we pilot generative AI tools, we are monitoring their efficacy in improving the findability of resources on SERC-hosted sites, retaining only those features that add value to visitors' experience.