Ga naar hoofdinhoud

Plain language summary

A publication can carry a summary in plain words next to its own summary. A Woo decision is written for lawyers. The plain language summary tells every reader what the decision means for them, at a reading level the officer names.

The fields​

FieldWhat it holds
plainSummaryThe summary in plain words, at most 1,500 characters.
plainSummaryLevelThe reading level it is written at: A2, B1 or B2. The default is B1.
plainSummaryProvenanceWho wrote it, with which text assistant, when it was accepted and by whom.

Taalniveau B1 is the level the Dutch government uses for text written for citizens. Choose A2 for text that must reach readers with low literacy. Choose B2 only when the subject cannot be explained at B1. The level is what you declare: the app does not measure the reading level of a text.

A summary without a level is refused, and the error names plainSummaryLevel. A summary saved without a level field at all gets B1.

The provenance is written by the server. A save that changes the summary records the person who saved it. A save that tries to write the provenance itself gets the stored provenance back. Anonymous readers do not see the provenance, because it names a user.

Writing it​

Open the publication. The Publication card shows the plain language summary and its reading level next to the summary. Edit both there and save.

Some tips for B1:

  • Use short sentences, with one message per sentence.
  • Address the reader as "u" and say what happened to them.
  • Use common words. Explain a legal term when you cannot avoid it.
  • Say what the reader can do next.

For example: "U kreeg geen parkeervergunning. U maakte bezwaar. De gemeente blijft bij het besluit."

Drafting it with a text assistant​

When a text assistant for simplifying text is installed in Nextcloud (hermiq, or any other app that offers the task type core:text2text:simplification), an officer can ask for a draft. The assistant gets the title, the summary and the description of the publication.

A draft is a suggestion. Nothing is stored until the officer edits it and accepts it. The stored provenance then reads ai, names the text assistant and names the officer who accepted it. A draft that fails or times out stores nothing, and the officer sees the error.

Only an officer who may change the publication can ask for a draft. Without a text assistant, writing the summary by hand always works.

RouteWhat it does
GET /api/plain-summary/assistantAnswers {available}: whether a text assistant can draft.
POST /api/publications/{id}/plain-summary-draftAsks for a draft and answers {taskId}. Answers 409 no-provider without a text assistant.
GET /api/publications/{id}/plain-summary-draft/{taskId}Answers {status, text?, error?}. Stores nothing.
POST /api/publications/{id}/plain-summaryStores {text, level, fromTaskId?} with its provenance.

On the public API and in DCAT​

The public publication response carries plainSummary and plainSummaryLevel. The DCAT feed emits the plain summary as dct:abstract, tagged with the publication's language, or nl when the publication has none.