If you click on the "Model" tab in the project overview you will see a page that with the following tabs:
- Document - description of data in your model, presented as form fields
- Structure - nested groups of fields from your document
- Style - Theming options for annotations on images
- Preview - where you can test your model
There are 2 main sections of a model. The first is the document. Each image you crowdsource will have a copy of this document. All of the contributions from users on the same image will be building up this single document. A document can have either fields for values inputted, such as transcriptions fields, or Entities. Entities are a way of nesting related fields.
A contributor may be able to add multiple instances of each of these fields and entities. Your document may contain a lot of different fields that could be crowdsourced. To manage this and avoid a single large form being presented to the user the document is split up using a Structure. A structure is a subset of the fields from the document. This drives a navigation so the user can decide what they want to contribute and then only see the fields for that thing.
A capture model is a digital document that is presented to a user as a web form. Capture models are like digital paper forms. Once created they are digitally "photocopied" and given to users to fill out. For each image that is annotated there will be a pile of these filled out forms from various users. These are consolidated into a single form.
This metaphor fits the goals of the capture model format well. When someone contributes the data is stable. Each document not only describes the data, but how to correctly edit the data - just like a physical paper form. Luckily as these are digital we can do a lot more with the data.
Each image has an associated capture model for each project it is in.
A capture model is composed of 2 primary pieces of information:
- Document - The data and how to edit it
- Structure - Subsets of the document grouped together into chunks easier for filling out.
The document should mirror how you would like the data to appear at the end of this whole process (e.g. in JSON) and the structure will drive how users are presented with that data. The structure allows you to give users a choice of what information they want to enter, perhaps based on what they are looking at in the image. This choice will drive which fields in the document are shown. At the end of the day though, regardless of the choice they make, they are still contributing to a single document.
An important aspect of capture models is the revision. Each time a user fills out some fields in a document, these fields are grouped together as a revision by that user. When a revision is created it does not become part of the canonical document until it is approved in a review. How this works is described in more detail in the Reviews section.