← MapEtyMap

Editorial Policy

How content is created, how automation is used and how uncertainty is communicated.

How content is created

  • Manual research

    Individual etymologies researched against published place-name dictionaries and historical records. Marked Researched or Source-backed.

  • Automated generation

    A pipeline using a language model (Gemini 2.5) generates candidate interpretations where no manually verified record exists. Marked Automated or Provisional.

  • Lexicon analysis

    A structured lexicon of place-name elements provides token-level analysis, supplementing both manual and automated records.

AI and automation

Automated output from language models is not presented as verified fact. All AI-generated records are:

  • clearly labelled as automated
  • assigned a lower confidence level
  • subject to lower quality thresholds for search indexing
  • not awarded the same prominence as source-backed records

The use of AI is a practical measure for covering a large dataset at low cost. It does not substitute for specialist scholarship.

Source selection

  • Scholarly credibility

    Preference for peer-reviewed works and publications from recognised place-name organisations (e.g. the English Place-Name Society).

  • Licence compatibility

    Only sources with a compatible licence are incorporated. Sources with unclear or non-commercial-only licences are not used until confirmed.

  • Traceable provenance

    Each source is identified by name and, where possible, by a specific page reference.

Communicating uncertainty

  • Documented — Supported by published sources with historical attestation.
  • Probable — Consistent with known patterns; not fully attested.
  • Suggested — Plausible but limited evidence.
  • Uncertain — Origin unknown.
  • Automated — Generated by software; not individually verified.
  • Provisional — Awaiting review.

Researched vs provisional records

Researched

Reviewed by the editor against published sources. May be promoted to this status after review.

Provisional

Output of automated processing not yet reviewed. Published as useful first-pass interpretations but clearly distinguished from researched records.

How updates are made

Content is updated in batches as the editorial pipeline runs or as corrections are reviewed. Material changes to attested records will reference the source that prompted the change.

New data processed by the automated pipeline is ingested periodically. Automated records may be regenerated as the pipeline is improved.

Contact

Questions about editorial decisions can be directed to the contact page. To report an error in a specific place record, use the corrections process.