Internet2 Mtg- Univeristy of Delaware Knowledge Graph Tool 9 Jan…
Look on this page for attached presentation: https://spaces.at.internet2.edu/display/CAI/2024-2025+Presentationspresentation from Erin Sicuranza’s team (mostly Jevonia Harris and Joe Naccarato)Pillars of effort:AIPrecise dataFaculty expertiseArchitecture benefits:Easy integration of new data sourcesSome stuff I missedPrompt RepositoryPrompts are broken down into parts and re-arranged for testing and re-useTranscriptsKaltura provides transcription and API access to let us gather and filter contentWorkflowPrompt DevelopmentPrompts are then combined into chain of thought style packages that run over all transcripts for a course/faculty comboFaculty Next StepsThey can now prompt against their own data via a topic development interfaceKnowledge GraphConnects faculty verified course data to faculty reviewed study aidsRelationship data can be presented to students and faculty to improve course materials and study habitsUsed Neo4J and LangchainCourse GraphCourses organized by subject and facultyCourses contain materials and material chunksTopics are connected to material chunksLearning object nodes are connected to a cousrseMaterialChunk and topicFaculty InterfaceNested course structureTopics and sub-topicsWith definitionsContext etccan set any given topic’s location in priority (main versus sub-topic)Student interfaceCan build and customize study decksSelect which cards to include, which notnext steps will be on study sessions and graph views of how they best study