eye-2771174_960_720

Better AI Retrieval Starts with a Better Taxonomy

When an artificial intelligence (AI) assistant answers a question, the quality of its response depends on what it retrieves. Ask the assistant for the current policy on research data sharing and a system might surface a superseded policy, a draft, or guidance meant for a different institution. All three can look relevant. Precision means finding the one that actually applies. This important and timely topic came to us from the National Institutes of Health and their article, “The Semantic Precision for AI Retrieval of Knowledge (SPARK) PubMed/PMC Challenge.”

Semantic technology helps AI interpret meaning and context. It can connect different ways of asking the same question and distinguish between identical words used in different contexts. In a university collection, for example, “IR” might refer to an institutional repository or institutional research. A system that recognizes both meanings can use the rest of the query to narrow its search.

A solid taxonomy gives that system a shared vocabulary for classifying content. Preferred terms, synonyms and relationships make it easier to retrieve records even when users and authors use different language. This matters because a fluent answer can hide a retrieval error. If the system pulls the wrong policy, it may summarize that policy accurately and still give the user bad guidance. Testing retrieval against real questions reveals where terminology, tagging or missing metadata sends the system to the wrong source.

Taxonomy is part of the infrastructure that makes AI answers dependable. When organizations invest in the language and structure of their information, they give AI a better chance of finding the right source and users a clearer way to verify the answer.

Without semantic enrichment, AI systems can retrieve text but not truth.

Stay current on the ideas shaping data and AI. Subscribe to Taxodiary for insights from the Access Innovations team.

Melody K. Smith

Data Harmony is an award-winning semantic suite that leverages explainable AI.

Sponsored by Access Innovations, the intelligence and the technology behind world-class explainable AI solutions.

Posted in

Melody Smith

Melody K. Smith has provided organizational, social media and digital communication services to a large non-profit for the past twenty years. Prior to that she championed employee engagement and communications in the healthcare field. She holds a Bachelors degree in Marketing. When not wrangling and writing TaxoDiary content for your reading pleasure, Melody writes fiction, rescues dogs and throws legendary dinner parties.

Leave a Comment