Data model (ontology) for Data Basket
We try to reuse established ontologies as much as possible. We considered using the latest version of Open Annotation (OA) as a basis, but gave up on this idea.In order to have a simple and coherent model, Data Basket uses the same basis as Data+Image Annotation: Open Annotation Collaboration (OAC). Still, we use some approaches from OA, eg ore:Aggregation.
We map entities from the Data Basket Spec 3.5 to the following:
|Basket||List of Bookmarks, owned by a particular user||ore:Aggregation|
|Bookmark||Pointer to an Item, made by particular user, with optional description||rso:Bookmark|
|Target Item||RS URI or web URL or pure text, (in some cases) having Preview, short URL, creator||see Item Types|
- Jana: why do we need a Basket?
- Vlado1: To distinguish the bookmarks of one user from those of another
- Jana: I rather imagined it denormalized with owning user being just a property of the bookmark. Having a basket is not a problem of course - just that we never actually deal with this object and we need some additional effort for maintaining it.
- Vlado1: To distinguish the bookmarks of one user from those of another
- We use rso:Bookmark's (a subclass of oac:Annotation) to easily distinguish bookmarks from other annotations, eg forum posts.
- Bookmarks can also be distinguished by the fact they are in a Basket. We never search for "all annotations", we search for annotations related to something an Object, or in a Basket, etc
- OA has such class (oax:Bookmark), but they still oscillate between using types and using a separate field called oa:hasMotivation
- Jana: We need to double check this won't break the existing Annotation API. I suggest someone reads through all SPARQL queries related to annotations.
We use the following prefixes:
|Pref||URI||Get from URL||Local file|
|ore||http://www.openarchives.org/ore/terms/||defined one inverse in rso.ttl|
A Basket is a bag of Bookmarks (represented as ore:Aggregation), owned by a particular user. Fields:
- A basket is created when the first bookmark is being added (on demand)
- A basket is never deleted, even on action Clear All
- For now users are represented with a Nuxeo username, not with URI
- TODO Jana: Can the given URI patterns be implemented?
Target Items can have one of the following types. The type is stored in two places for convenience:
- rdf:type of the target.
- rso:hasTargetType of the bookmark. This custom property stores exactly one of the types below, while the above also stores all supertypes.
|Data Annotation||rso:DataAnnotation|| see Annotation with OAC and Reification
New subclass of oac:Annotation that allows the Basket code to recognize more easily the item type.
TODO Mitac & Stanislav: Change the Data & Image Annotation backend to use rso:DataAnnotation and rso:ImageAnnotation
|Forum||Not in RS3.5|
|Image||crm:E38_Image|| Dominic: we don't have a display of the image only, so do we need this? Yes, we will need this soon
TODO Vlado: ##Note2: change RKD mapping to make explicit image URL instead of has_image_file and attach E38. BM mapping already does that
|Image Annotation||rso:ImageAnnotation|| see Image Annotation Design#Example.
New subclass of oac:Annotation: see comment of Data Annotation
|Object||rso:FC70_Thing||This class is set by a rule in FR Implementation|
|Object Field||rdf:Statement|| see Annotation with OAC and Reification.
When an AP is added to Basket, this Statement is created, same as making annotation or a link to AP
|Search||rso:Search||new class. It has one field P3_has_note storing the textual representation of the search.|
|Text Snippet||(none)|| Does not have item URI, therefore no type. Its title and content (rich-text description) are saved in Bookmark's body
TODO Svetoslav & Stanislav: assume that no type means Text Snippet
|Web Link||foaf:Document|| New class with no fields.
TODO Svetoslav: add statement as per ##Note1
- Vlado: should the rso:Search node be created when the Search is put in history, or only when it's put in basket?
- Jana: History is currently stored in Nuxeo
- Vlado1: I understand, but I thought we'd be storing the history same as the bookmarks? The question is who & when will create the node
- Jana: We be better store history in RDF indeed using the same data model. Still, it should be distinguishable from bookmarks as we need them separately in UI. (We may use a second basket for the user - a history one). I think we should use rso:Search for storing queries from history as well as saved searches (when we implement these).
- Vlado2: doesn't "saved search" mean the same as "search placed in basket"? The Search node is the same, no matter whether it's in one or more Histories and/or Baskets
Each bookmark resides in a basket and points to a target item (URL). Fields:
|UI Field||Property (path)||Applicability and Notes|
|rdf:type||rso:Bookmark (fixed value)|
|Type||rso:hasTargetType, oac:hasTarget/rdf:type||All except Text Snippet. See Item Types|
|Preview||rso:hasTargetPreview||Image/Image Annotation: RS image URL. Web Link: URL from Web Preview service|
|Title||oac:hasBody/rso:P3_has_title||Plain text. Copied from Item and shortened to 30 chars|
|Notes||oac:hasBody/rso:P3_has_description||Rich text. No default value|
|URL (Item)||oac:hasTarget||All except Text Snippet|
|Short URL||rso:hasTargetShortURL||From URL Shortening service|
|Item created by||oac:hasTarget/dcterms:creator||Only for Data/Image Annotation|
|Item created on||oac:hasTarget/dcterms:created||Only for Data/Image Annotation|
- Applicability describes for which item types that field is applicable, and gives extra notes
- The following fields shown on Data Basket UI design don't really exist: Sent by, Tags, Last updated
- oac:hasBody/dcterms:creator&created are set from the current user and datetime when the bookmark is created.
They are preserved if the bookmark is copied to another basket.
- oac:hasTarget/dcterms:creator&created are present only for 2 item types. They alraedy exist, they are not added when the bookmark is created.
In the illustrations below, a red square indicates a pre-existing statement. All others are created when a bookmark is created.
See Image Annotation Design#Example (we present only the relevant part of the image annotation's data)
- We omit http://www.researchspace.org/ from the URLs by using @base.
- The hasTargetPreview URL should be resolvable! It's given above as <DT219363.tif>, but an appropriate thumbnail-size IIPserver URL should be used instead
- dcterms:creator&created from <DT219363.tif/annot/1/body> are displayed in the Bookmark, although they are not stored as part of the bookmark
- The illustration below shows "DT219363:tif" instead of "DT219363.tif" because of limitations in plantuml (treats "." as a package name separator)
- It's made with plantuml.jar standalone because of incomplete installation: OA-1608
- The specification is written specific according to existing tools. However, the way that the data basket works should be generic for existing and new tools to come.
- The production version of ResearchSpace will require an API that means that all tools that adhere to it can be accessed and inetgrated with the databasket tool. This means that there must be a consistent type of URI / URL with tool parameter so that tools can be launched and placed in the appropriate state.
- These links should be ones that could be used as a normal web link in a browser (RESTful) so that the links could be used outside the databasket using a simple browser address box.
- To this end the specification should also have a technical specification outlining the way in which the databasket would interact with the ResearchSpace environment
- I agree that bookmarking should be more in the style of using "browser address box". However, our URIs are currently not resolvable, see RS URLs and URIs.
Adding a new item type ("tool") to the basket is a matter of adding it to the table above, and to frontend handling code (different types have somewhat different handling behavior).
- For full extensibility the Basket should be able to figure out the item's type from the item URI alone. That is the case above, except for Web Link (for which we need to add a type statement upon adding to basket).
- I've used "a foaf:Document" above (that's the range of foaf:homePage)
- OA recommends using "a dcmitype:Text; dc:format "text/html" which I don't find very nice (and may change)