Universidad Continental: artificial intelligence for metadata and cataloging on a commercial system already in place

Institutional context

Universidad Continental is a private Peruvian university with multi-campus operations and its main campus in Huancayo. Its information hub runs on a commercial library management platform, a decision made and implemented beforehand that was not up for discussion.

The challenge

The university's position was the one we hear most often, and the one most vendors sidestep: we have the system we want to have; what we need is for it to stop costing us so many hours of keying. Automating document description is usually sold tied to a platform: you get automation if you use my system. That turns a process improvement into a migration project, with its cost, its risk and its year of execution.

The technical challenge, then, was not cataloging with artificial intelligence, but doing so on a third party's platform, respecting its integration rules and without touching the workflow the team had already internalized.

What we implemented

• MetaExtract API for the automatic extraction of metadata from documents and the auto-filling of descriptive fields.
• BiblioScan AI for AI-assisted cataloging of physical material from photos, with auto-filling of MARC21 fields.

Both tools operate as a layer on top of the existing platform: they do not replace it, they do not duplicate it and they do not require the university to maintain two systems in parallel.

How we did it

1. Assessment of the target environment. We reviewed the metadata schema in use, the institution's cataloging profile and the integration paths the commercial platform allows.
2. Tuning to fields and document types. Extraction was trained with real material from the university, so that the output would match its profile and not a generic standard.
3. Integration without altering the workflow. The tools were connected so that the result reached the team's working form, not an intermediate file that someone would have to upload again.
4. Validation with a control batch. A representative set was processed and compared with the work of the team itself, which is the only valid baseline.
5. Training and staged rollout. First MetaExtract API for document description and then, once that service was stable, BiblioScan AI for cataloging physical material.

Result

The university gained automation capacity without giving up its earlier investment in a platform and without taking on the risk of a migration. The two services were contracted at different times and both have been renewed, which indicates that each one was evaluated separately and the decision was made to keep it. The relationship has since expanded toward other components of our portfolio, which the institution is currently evaluating.

Technologies

MetaExtract API, BiblioScan AI, integration on the existing commercial platform, MARC21, metadata extraction with language models and image recognition applied to cataloging.