What MetaExtract API is
MetaExtract API is an application programming interface developed by BiblioLabTech that reads a PDF and returns its metadata already structured: title, authors, abstract, keywords and the additional fields your institution needs, ready to pre-fill the description form of a repository or catalog. It has no screen of its own. It plugs into the system you already run and takes transcription out of the description workflow.
What separates it from generic text extraction is interpretation. MetaExtract API uses language models to understand the role each fragment of text plays in an academic or institutional document: it tells the title from the subtitle, separates author names from their affiliations, and finds where an abstract begins and ends even when the document does not label it.
Which fields it returns
The field set is trained to your schema, and in practice it consistently covers:
- Title and, when present, subtitle
- Authors, with their institutional affiliation when the document states it
- Advisor or supervisor, for theses
- Abstract, located in the body even when it is not marked as a section
- Keywords, both those declared by the author and suggestions drawn from the content
- Date of publication or deposit
- Document type, such as thesis, article, report or conference paper, useful for routing items to the right collection
- Language of the document, extent and identifiers
- Local fields your institution defines: academic program, academic unit, project number, funding call or any other element of your schema
That last point is what separates a generic extraction from a useful one: the output corresponds to your schema, not to a standard template that someone has to rearrange by hand.
How the integration works
- Model training. The extraction is trained on your institution's real theses, articles and reports so it recognizes your document types and local fields, and it is connected to the target repository or system.
- Document upload. The PDF reaches the API either when someone uploads it to the deposit workflow or automatically when your system receives a deposit.
- Recognition and categorization. The API identifies the requested metadata and organizes it according to the agreed schema: Dublin Core, local fields or both.
- Field pre-filling. The extracted data fills the description form of the target system, ready for review before publication, with no intermediate file that someone has to reload by hand.
Our guide to AI metadata extraction covers each stage in more depth.
Accuracy and human review
No automatic extraction is right on every document, and BiblioLabTech does not promise that. Fields written explicitly in the document, such as title, abstract or date, are extracted with high reliability; fields that require interpretation, such as thesaurus descriptors or the license, always need review. Born-digital PDFs with selectable text give better results than low-resolution scans or unusual layouts.
That is why every workflow includes human review before publication. The librarian applies judgment where the machine cannot decide: name normalization against the institutional authority file, thesaurus descriptors, and license and access decisions. The aim is not to remove review but to confine it to the cases that need it.
Which systems it works with
| System | Typical use of MetaExtract API |
|---|---|
| DSpace | Pre-filling metadata in the institutional repository's deposit workflow |
| Koha | Describing digital documents linked to the catalog |
| ABCD | Extracting metadata for catalogs built on this platform |
| OJS | Describing articles in scholarly journal workflows |
| Commercial library platforms | Integration through the routes the vendor enables, without migrating |
Multi-platform compatibility solves the hardest scenario to automate: a library that runs two platforms, such as a repository and a catalog, and needs to reduce transcription in both without merging them into one system. Every DSpace implementation by BiblioLabTech includes MetaExtract API at no extra cost, and the API is also licensed on its own for existing repositories, including ones we did not build.
Who it is for
- Institutional repositories with a steady flow of deposits, where theses, articles and reports arrive every semester and manual description becomes a bottleneck that grows each term.
- Libraries with a digital backlog: digitized collections and historical institutional output that were never described for lack of staff. Batch processing handles them by collection or document type, reviewing each batch before moving to the next.
- Institutions running more than one platform, such as a repository and a catalog, or a repository and a journal system, where one trainable extraction replaces two manual workflows.
It does not make sense when the volume of documents is low and sporadic: training and maintaining an integration only pays off with a constant flow of material, and below that threshold manual description remains reasonable. Our overview of the institutional repository explains why so many repositories launch nearly empty and stay that way.
MetaExtract API or BiblioScan AI?
Both automate description with AI, but they start from different sources. MetaExtract API reads digital documents in PDF and is built for repositories and document description. BiblioScan AI reads photos of print material and is built for cataloging physical collections in MARC21. Libraries with a print catalog and a digital repository often run both, one for each front of the collection.
Timeline and pricing
Implementation takes six to nine weeks from diagnosis to production, and less when your metadata schema is already defined and documented. MetaExtract API is licensed as an annual service that includes the initial implementation, adjustment to your fields, support and improvements during the period. The price depends on the number of schemas or document types to support, the integrations required and the expected processing volume. No infrastructure is needed on your side, because it is an API service.
Where MetaExtract API runs in production
- Universidad ICESI (Colombia) added MetaExtract API to its existing description workflow without changing systems or migrating content, and has renewed the service for consecutive terms. Read the case study.
- Universidad Católica Luis Amigó (Colombia) runs it on two platforms at once, its DSpace repository and its ABCD catalog, continuously since October 2024, with the contract renewed. Read the case study.
- Universidad Continental (Peru) runs it on top of a commercial library platform, without migrating. Read the case study.
- Universidad Americana (Paraguay) uses it in the deposit workflow of its DSpace repository, hosted by BiblioLabTech. Read the case study.
- Escuela Colombiana de Ingeniería Julio Garavito (Colombia) uses it in its DSpace deposit form, with mandatory human review before each record is saved. Read the case study.
Different institutions, different combinations of systems, and the same pattern underneath: the API adapts to what already exists. Tell us about your repository and your metadata schema to find out how quickly the extraction can be trained on your material.