Data Policy and Copyright


Data interventions

Several types of interventions (marked with an ⓘ symbol in object records) have been made to the original data submitted by museums, in order to make it easier to access and use within this national portal.  

Interventions include:

  • Data reconciliation
  • Data generation

Data reconciliation 

Fields where original data from museums has been be mapped to established terminology standards, or otherwise standardized, in order to enable faceting, sorting, and other useful ways of accessing and displaying the data:  

Object Classification 

An automated pipeline that standardizes object type terminology at scale using two services:

  • The Nomenclature for Museum Cataloguing reconciliation API which is the industry standard controlled vocabulary for naming museum objects
  • Anthropic's Claude AI (via the Batch API) which is to validate and suggest the best term when API results are uncertain or missing

How the Pipeline Works

  1. Nomenclature API Matching: Each object's title or type is sent to the Nomenclature reconciliation service, which returns the closest matching official term and a confidence score.
  2. Prepare AI Batch: Records are grouped by unique combinations to avoid sending duplicate requests. Only one Claude request is sent per unique combination.
  3. Submit to Claude: The batch is submitted to the Claude Batch API for asynchronous processing.
  4. Wait for Completion: The pipeline polls automatically until the batch is done.
  5. Download & Validate: Claude's suggestions are validated against the Nomenclature vocabulary. Each record is marked as validated or flagged for human review.

Data generation

Data that has been added or assigned to improve access or clarity

  • Discipline (“Humanities” or “Natural Sciences”) has been added to allow filtering of object records.
    • Records that have Specimen Taxonomy will be assigned as "Natural Sciences".

Data transparency

Even though these interventions are displayed alongside data from museums' records, Histellis will always be transparent about what is original data submitted by museums, and what has been added or enhanced through A.I. or manual processes. 

Language of data

Although the Histellis Collections Portal website is (will be) available in both English and French, collections information provided by museums may be available in one language only; Histellis presents museum information in the language in which it was provided.  

Content disclaimer

The information in Histellis collections portal comes from the collections databases of museums, and while this information is continually being improved and updated, in some cases it may be incomplete or in need of refinement. Museum collections documentation often has been developed over many decades through the contributions of staff and volunteers at museums and cultural centers. Keywords, titles, and descriptions given by the creators of artworks and museum workers are products of their time. Although museums strive to use respectful terminology and accurate representations, the data may sometimes include outdated or offensive terms. If you have concerns about any information found in Histellis, we invite your input by contacting us.

Copyright 


Copyright of data/media files belongs to the contributing museums. Copyright on the compilation (Histellis Collections Portal website) belongs to Canadian Heritage Information Network. 

Data use

Histellis data can be used in several ways:

  • Individual records within Histellis can be downloaded or shared, through the "download" button on each record.
  • API for Histellis data (future)  supporting multiple output formats (JSON, XML, CSV).

The data shown in Histellis records are a sub-set of the museums' information on the objects; for further information on the object or the record, contact the museum that contributed the record to Histellis

Please ensure that your intended use of these records is permitted by the data-use licences stated within each record as applied by the contributing museums.  You can refine your search using the data use licences filter if needed.