. . . . "Is all research worth considering? Even with a doubtful yes, the question becomes how to gain a holistic view of 180,000 Parkinson's disease (PD)-related papers. Large language models can read them, but they answer in a stochastic way, as they were trained. Knowledge graphs are a method to force the LLM to consider deterministic relations before answering. But constructing meaningful KGs by human effort is limited — as was our goal, twenty years ago, of creating a semantic web based on ontologies.\r\n\r\nsciai-lab constructs the ontology from the data it extracts from the literature. The first round used a simple schema that showed the LLM what to extract; by studying the extractions, each successive schema became more sophisticated. In the meantime we have built an ontology (v0.2, on BioPortal) with 21 entity types, 29 relation types and more than one million assertions. Every assertion is a femto-publication. We don't ask the researcher to change their ways. We are retrofitting the prose." . . . "Johannes Keizer: \"sciai-lab: ontology-guided AI to mobilize research data\"" . . "Virginia Balseiro" . "2026-10-12T13:40:48.677Z"^^ . . . . . "Agenda item: Johannes Keizer: \"sciai-lab: ontology-guided AI to mobilize research data\"" . . . . . "RSA" . "MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQCFJNRSo0AhDh7EfwM3nZXQbACb8v6F7tKGOj4Mnc/VuEu0CqzwyomaSvXmfwIKeHmCGCdIrL7tMes3U3K7qJ6c3m5j9U1SDBA+d6UDGvBKSN4X8vvRHzH+PNZyegn3Wu+liXjq+4bnGdTdhPRdRFO9DjSb+rpAfaH21md4qRhCewIDAQAB" . "GICoOPjLT85S21m6cYemZt8WNYyDZIklsc1OJGgch3Xdwo1I8ZaetFplTtiVhAfys4iJY77yoCBUDzjFc6jW5V3bnn6Ke479/UDWylWURrgKENFE3TdtX/JqI0MpSfMlEFjKf9y2zeMEboI5cxHqvEk8tSSTg6hN11cExFQq3yI=" . . . "Nano Session #35" .