See also: IRC log
<trackbot> Date: 01 October 2014
<fjh> Agenda: http://lists.w3.org/Archives/Public/public-annotation/2014Sep/0075.html
<azaroth> and for making notes, taking "votes" and so on
fjh: TPAC is coming up -- deadline Oct 8th
... make sure you've made arrangements
... have also sent out a notice about an upcoming privacy workshop that
may be of interest to this group
<fjh> privacy workshop notice http://lists.w3.org/Archives/Public/public-annotation/2014Sep/0072.html
fjh: i think we decided on the list to use
GitHub for issue tracking
... but there are questions about how to get notifications on the list
<bigbluehat> Zapier.com + mailing list posting?
<bigbluehat> Webhooks (broadly)
fjh: any suggestions for how to do this?
<bigbluehat> +1
<ivan> +1
<Jacob> +1
<fjh> proposed RESOLUTION: use github for issues tracking, assuming we will get notices to public list working later
ivan: if you have a good trick for this, let me know off-list
RESOLUTION: use github for issues tracking, assuming we will get notices to public list working later
<fjh> proposed RESOLUTION: use tracker for actions
<ivan> +1
RESOLUTION: use tracker for actions
<fjh> we may use tracker to capture issues quickly during a call but will then use github for full issue tracking
<fjh> proposed RESOLUTION: Minutes from 24 Sept 2014 approved, http://www.w3.org/2014/09/24-annotation-minutes.html
RESOLUTION: Minutes from 24 Sept 2014 approved, http://www.w3.org/2014/09/24-annotation-minutes.html
<azaroth> http://www.slideshare.net/azaroth42/open-annotation-core-data-model-tutorial
fjh: if you want an issue created, put it in irc with "issue: ..."
<fjh> http://www.slideshare.net/azaroth42/open-annotation-core-data-model-tutorial
<fjh> Highlight known issues, WG members indicate additional issues, discuss
azaroth: not everyone has come from the community group so we're going to work through the data model and review people's questions and concerns
<tilgovi> tilgovi
azaroth: link in IRC is a slideshare
presentation about the core data model
... please shout if you have questions as we go through this
... first slide: what is annotation?
... one of the issues we had in the community group: how to define what
an annotation is?
... Annotation: a set of connected resources, typically including a body
and target
... example use cases: as aide-memoire, share, improve discovery,
organise, interact with other people, being a creator (e.g. editor) or
moderator ... over the linked resources
... that's the baseline scope that was considered in the CG
<fjh> slide 4 core data model
azaroth: which resulted in our baseline
model for annotations: a separation of the annotation, which
encapsulates the linkage between body and target
... a freq question is "how do I give the relationship between body and
target?"
... we don't want to reinvent RDF
<fjh> slide 5
<fjh> now slide 6
azaroth: example here on slide 6 --- a video
about a Hubble Deep Field image
... [the body is implicitly related to the target]
... we're on slide 8
<bigbluehat> this deck, yes? http://www.slideshare.net/azaroth42/open-annotation-core-data-model-tutorial
azaroth: useful to know the general type of body and target
<bigbluehat> dwhly: just wanted that in the logs :)
azaroth: e.g. body is text serves as a hint
to viewer about how to render the annotation
... versus audio, video, etc.
... also useful to know the media type
<JakeHart> JakeHart is IPcaller.a
azaroth: which might be codec, etc.
... recommendation to use Dublin Core types vocab
... on slide 9
... body has dc:format of "application/flv" and rdf:type of
"dctypes:MovingImage"
... target similarly has dc:format and rdf:type
... slide 11
... embedded bodies -- body can be an inlined resource as an alternative
to a resource with a URI
... concerns that people would never implement this
... slide 12
... this was a hotly debated point within CG
... slide 13
... settled on W3C's Content in RDF specification
... seemed at the time like it was going forward towards full TR or some
way there
<fjh> http://www.w3.org/TR/Content-in-RDF10/
azaroth: but that hasn't happened as yet
... so we may need to discuss this within the context of this Working
Group
<ivan> ISSUE: what do we do with embedded content
<trackbot> Created ISSUE-1 - What do we do with embedded content. Please complete additional details at <http://www.w3.org/annotation/track/issues/1/edit>.
azaroth: do we want to stick with this or change the way we do embedding?
<fjh> ISSUE-1: how to embedd content in graph, given status of Content in RDF spec, use that or change
<trackbot> Notes added to ISSUE-1 What do we do with embedded content.
azaroth: slide 14
<fjh> ISSUE-1: http://www.w3.org/TR/Content-in-RDF10/
<trackbot> Notes added to ISSUE-1 What do we do with embedded content.
azaroth: embedded body has rdf:type ->
"cnt:ContentAsText"
... slide 15
... tagging -- the simplest form of annotation
... comes in two forms: 1) plain text tags 2) semantic tags (tags with
URI that *identifies* the concept)
... also the ability to tag with a document, standing for a concept
... slide 16
... we give this rdf:type "oa:Tag" to distinguish the intent of this
kind of body
... slide 18
<MGU> Question: Is not oa:Tag redundant with motivation predicates ?
azaroth: semantic tagging is even easier,
has rdf:type "oa:SemanticTag"
... and the URI refers to the concept-as-tag
... there was some consensus that this was the most appropriate, if not
perfect, way of doing this
... we didn't think there was anything wrong with tagging with documents
... but this might be an issue for re-discussing
... in order to avoid polluting resources as a result of being tagged
MGU: is there a relationship between tagging and the motivations part of the model
<ivan> ISSUE: with semantic tags, how to avoid polluting the space (eg, dbpedia Astronomy) with annotations
<trackbot> Created ISSUE-2 - With semantic tags, how to avoid polluting the space (eg, dbpedia astronomy) with annotations. Please complete additional details at <http://www.w3.org/annotation/track/issues/2/edit>.
azaroth: yes, in that any annotation which
has a tag body should probably have "tagging" as one of its motivations
... it's probably slightly redundant, but in an annotation with non-tag
content, you need to be able to distinguish between the tags and the
non-tag content
paoloC: want to reinforce this point -- can have annotations with a motivation of "commenting" that nonetheless have tags
<fjh> ISSUE-2: motivation can be commenting with comment, even if a tag is added , example in biology case might want to comment but tag t clarify scope of comment for example
<trackbot> Notes added to ISSUE-2 With semantic tags, how to avoid polluting the space (eg, dbpedia astronomy) with annotations.
paoloC: in biology, we may have annotations with a lot of text that references a lot of biological items/artefacts, which might be added as tags, even if the original motivation was commenting
azaroth: i don't think we want to discuss
solutions at this point -- we'll be jumping all over the stack
... but do jump in with questions or issues
... slide 20
... we found a lot of systems want to use a document in place of a
semantic tag, e.g. using a Wikipedia page to represent the content it
describes
... in order to avoid identifying the document as a semantic tag (which
would be very wrong)
... we introduce a new object between the document and the annotation,
which has type "oa:SemanticTag" with a foaf:page relationship to the
document
... this object can be empty
... slide 21
... but it could be an object with, say, a UUID
... slide 22
... most *non-tag* annotations are about *segments* of a resource
... (unlike "+1"s and "likes" and so on)
... use URI fragments to identify and describe segments
... Media Fragments WG have developed ways to identify fragments of
images/video/etc.
... great for simple cases but not sufficient
... slide 23
... can use these Fragment URIs as targets
... slide 24
... targets can reference these URIs, such as http://example.com/myimage.jpg#xywh=100,500,200,150
... can also use same trick with bodies
... slide 25
... for example if only a short section of video annotates a particular
resource
... slide 26
... model allows for annotations without a body
... for use in, e.g., bookmarking and highlighting
... slide 27
... you're *implicitly* saying "i like this" or "i want to come back to
this" -- it's not explicit in the model
... slide 28
... multiple bodies and targets: each body applies individually to each
target
... slide 29
... for example, if you have one body about two targets -- you are
implying that the body annotates *each* target
csillag: can I express if one body is annotating all the targets?
azaroth: we have notions of composites (set
of all these things) and lists (set of all these things with order)
... so you could indirect the target through a composite or list
... another potential issue is whether we need composites (sets) as well
as lists
<ivan> ISSUE: do we need both composites and lists
<trackbot> Created ISSUE-3 - Do we need both composites and lists. Please complete additional details at <http://www.w3.org/annotation/track/issues/3/edit>.
azaroth: is the distinction between ordered
and unordered sets of objects important?
... slide 30
... we felt it was important to record the provenance of the annotations
... useful for spam, reputation models, etc.
... when the annotation was made -- needed for synchronisation
... also needed to establish if the targets might have changed since the
annotation was made
... who or what serialized the annotation to the document format?
... this being useful for debugging, advertising, etc. (e.g. "Posted
from TweetDeck")
MGU: what about signing annotations with a
private key?
... is there a way that an author can sign an annotation to prove
authorship?
<ivan> ISSUE: do we need signature possibilities to annotations (not in the current OA model)
<trackbot> Created ISSUE-4 - Do we need signature possibilities to annotations (not in the current oa model). Please complete additional details at <http://www.w3.org/annotation/track/issues/4/edit>.
azaroth: we discussed annotation signing briefly, felt that it was quite complex and didn't want to add it as a burden [to implementers] if it wasn't necessary
<fjh> ISSUE-4: is there a requirement for integrity protecting annotations with digital signatures
<trackbot> Notes added to ISSUE-4 Do we need signature possibilities to annotations (not in the current oa model).
<Kyrce> +q
<Jacob> talking about roles for the annotation creator
<Jacob> i.e., creator was an editor, author, etc.
Kyrce: there is a linkage between the
annotation provenance and the role of the annotation creator
... will send an email to the list
<Jacob> IIRC, motivations were supposed to help support this.
<csillag> ISSUE-4: some previous discussion: https://github.com/hypothesis/h/issues/344
<trackbot> Notes added to ISSUE-4 Do we need signature possibilities to annotations (not in the current oa model).
azaroth: agreed -- the lack of a way to
record the role the creator is playing may be something we want to look
into?
... on the same resource the same person could have multiple roles at
different times
<fjh> ISSUE: define roles of actors, e.g. student, teacher etc.
<trackbot> Created ISSUE-5 - Define roles of actors, e.g. student, teacher etc.. Please complete additional details at <http://www.w3.org/annotation/track/issues/5/edit>.
paoloC: we intentionally didn't go beyond the provenance layer because when we discussed prov there were different ideas
<fjh> ISSUE-5: Kyrce will send email related to this
<trackbot> Notes added to ISSUE-5 Define roles of actors, e.g. student, teacher etc..
paoloC: we distinguish between who created
something, who edited, who curated, etc.
... all these aspects can be added to the annotation itself
<Gerben> +q
<fjh> ISSUE-5: paoloC notes that these things could be added to annotation itself
<trackbot> Notes added to ISSUE-5 Define roles of actors, e.g. student, teacher etc..
paoloC: we decided not to go into that level at the CG level
<MGU> I asked whether we would define a core annotatin system/voca and then extensions
<Gerben> fjh: yes
azaroth: in the CG we started off with the
core model and an extension set
... the issue we had with this was that it became impossible to know or
recall which things were in the core namespace and which were in the
extension namespace
... it became confusing why some things were extensions and some things
weren't
... in the current (CG) model, everything is in one namespace
... but that's not to say we can't do that here
... modules are a part of some W3C specifications (e.g. CSS)
... although I would want to defer any decision on that until the spec
became so large that it needed to be split
<fjh> ISSUE: are modules needed? , defer until we really need it with large spec
<trackbot> Created ISSUE-6 - Are modules needed? , defer until we really need it with large spec. Please complete additional details at <http://www.w3.org/annotation/track/issues/6/edit>.
<MGU> ok for me
azaroth: moving on... slide 31
... the model for provenance is that provenance data sits on the
annotation
... e.g. oa:annotatedBy, oa:serializedBy, oa:serializedAt, etc.
... slide 33
... there are different classes of agent (such as Person, Organisation,
Software)
... annotation model doesn't say anything about that: we leave it to
other data models (FOAF, etc.)
... slide 36
... motivations: we want to know why the annotation was created
... introduce a Motivation class, separate from Annotation class
... because we wanted to be able to use the richer SKOS Concept ontology
to provide cross-community resolution
... so another community can develop a different motivation, to which we
can link using SKOS
... slide 37
... a set of motivations from the specification
... such as commenting, bookmarking, highlighting, tagging, etc.
... slide 38
... we record this as oa:motivatedBy on the annotation
... that's the end of the core data model
... any more questions?
csillag: is there a list of the difference between bodies and targets -- what can bodies do vs targets?
azaroth: what do you mean by "do"?
csillag: what's the reason for the distinction?
<Jacob> Bodies and targets are symmetrical properties of the annotation but each has a different role with respect to the annotation.
azaroth: you need to know [which piece
annotates and which piece is annotated]
... but the objects have the same behaviour in RDF space
<fjh> big thanks to Nick for excellent scribing.
<paoloC> +1
<ivan> Thanks indeed!
<Jacob> +1
<MGU> thanks !
<ivan> and thanks to rob for the presentation
azaroth: please do bring up any issues and discussions on the mailing list
paoloC: question unrelated to the model --
during the week people said what they were interested in
... what's the plan for that data?
azaroth: we now have all but the last two or
three into my tracking spreadsheet
... it's reasonably well separated out
<fjh> +1 thanks to Rob for excellent presentation
azaroth: no orphaned topics, and none with
vastly more interest than others
... by the end of the week i will try to have that information public
FIN
<ivan> trackbot, end telcon