Provenance Incubator Group Teleconference -- 28 May 2010

<trackbot> Date: 28 May 2010

<YolandaG> scribe: Irini

<pgroth_> slides at http://www.slideshare.net/troncy/provenance-for-multimedia

<dgarijo> thanks

Slides from Raphael: http://www.slideshare.net/troncy/provenance-for-multimedia

provenance and multimedia

common scenario : mark & annotate

Semantic Web, instead of annotating with keywords, annotate with URIs both for properties and resources

In order to capture provenance, the naive approach uses RDF reification

Advanced approach: use COMM/M3O to capture provenance

can be recorded manually or be machine producible

complex RDF graphs can be produced from the provenance patterns

<pgroth_> not really...

<pgroth_> +q

Yolanda: is this the approach taken in the W3C multimedia group? what is the community represented?

Raphael: not W3C multimedia group. Prefer not to work with more complex patterns

community: multimedia community interested in semantics

Raphael: there are actions to bridge the approaches between simple and more complex models

Yolanda raises the issue of the mapping between the vocabularies

multimedia group: mappings between vocabularies when those overlap

two step approach: find the vocabularies that they overlap, talk to the authors to comprehend the meaning, and in the second step, provide a formal language, or an API, or a natural language explanation of the mappings. Multimedia annotation group will publish the set of mappings informally, and will request the formal mapping between the vocabularies.

Paul Groth: problem in provenance is how to "point" in complex objects to have fine grained provenance. Are there approaches for doing this?

Raphael: MPEG-7 is good for compound objects in multimedia
... result is the XML document that describes the decomposition introducing decomposition
... Media fragments working group deals with this to point directly to the fragment of the image/video/audio

<pgroth_> so the movement is towards using urls

Raphael: in the case of XML, you can use the URIs

<raphael> ... + XPATH expression

Yolanda: is there interest or is there work in describing the processes that are done on the multimedia content?

Raphael: important use case. The pattern approach works if one wants to model transformations on the image. It is important to track the changes operated on the images. News agencies: use the metadata of the photo to publish photos -- people change the metadata. they want to detect whether an image has been changed or altered in any way.

<pgroth_> +q

Paul: are there standards for embedding such metadata in images? JPEG has attribute value pairs. Are there standards for more structured metadata?

<pgroth_> p-

Raphael: No standards for storing this information

<pgroth_> xmp

<pgroth_> ?

<pgroth_> right

<pgroth_> yes you can

Ivan: are there tools to add metadata easily?

<pgroth_> python?

<raphael> http://www.w3.org/People/Bos/PhotoRDF/thumbnails-xmp

Raphael: yes possible: links is http://www.w3.org/People/Bos/PhotoRDF/thumbnails-xmp

(example link)

Yolanda: what is the state of the art utilization of technologies for verifying properties of the image

Raphael: exist declarative ways of representing securitty but can be embeded in the image. Multimedia community is based on the comparison of pixels of the image and the camera used to take the image (a kind of signature of the image). Also blobs are compared (8x8 pictures). Many techniques all rely on purely comparing pixels.

<pgroth_> very interesting

Yolanda: is there a possible connection with the provenance group and groups that Media Annotation Group ?

Raphael: yes, if we could attach provenance around the annotations of the picture
... has proposed the use of a Named Graph or a vocabulary
... it is not clear which is the appropriate use case.
... bring someone from the Media Annotation Group to talk to us about the vocabularies used and the mappings.

<pgroth_> do they know about our requirements doc?

Raphael proposed the chair of Media Annotation Group.

<pgroth_> I wonder if it reflects their requirements for provenance

Raphael: need for more communication to understand between the two groups.

(to establish an understanding)

<pgroth_> yeah not me :-)

Jose Manuel abt Internet Architecture trust and privacy

Jose Manuel: gave a talk abt the provenance and challenges w.r.t. trust and privacy

<YolandaG> Jose Manuel's email and pointers are at: http://lists.w3.org/Archives/Public/public-xg-prov/2010May/0009.html

interesting feedback regarding the role of provenance w.r.t. accountability and policies. Systems in the cloud take data from people and propagate data with policies that are not clear.

European Directive abt the use of private data.

Provenance can be used to establish accountability of systems w.r.t. existing policies.

Community is interested in attribution as well.

Interested to record who did what in a specific case and who is responsible for some actions in the Internet

In Future Internet assembly -- discussion on the Future Internet Infrastructure.

Important to monitor the movements of data in order to optimize the protocols for FI infrastructure

Positive Reaction -- Raphael was invited to Incotrust(?) workshop

<pgroth_> Jose Manuel

(thanks)

important to understand the processes used to manipulate the data

Invited R. Wright and J. Clark for a talk in the provenance inc. group Teleconf. Exists a potential to establish a collaboration with these security-related groups.

Jose Manuel: Incotrust workshop has participants from a number of countries to work on secutity/privacy/trust

Jose Manuel: people talk all the time abt provenance (attribution, accountability) w/out talking abt this specifically

Olaf: Abt Provenance Vocabulary mappings

<olaf> http://www.w3.org/2005/Incubator/prov/wiki/Provenance_Vocabulary_Mappings

If Common terms exist, use the provenance related OPM terms

If Common terms exist between vocabularies, use the provenance related OPM terms

for mappin terms use the skos vocabulary

wanting to enable gap analysis

How to proceed:

how can the mappings be described (as an RDF document?)

Is the exercise done? should we stop the exercise or based on the identified gaps increase the number of common terms?

<raphael> I think it is a good idea to have an RDF representation of the mappings

Yolanda: what did Olaf learnt from the exercise so far

Olaf: vocabularies he worked with are more specific than OPM has

Olaf worked with # prv: Provenance Vocabulary and [?]

[?] is# cs: Changeset Vocabulary

what is missing are concepts related to time from Changeset and not represented in the common terms here.

michaelp: OPM different approach in handling provenance from what the domain vocabularies have

Yolanda: exercise is helpful for us and others

<pgroth_> I had feedback today that it was very helpful

Yolanda: suggests adding documentation
... suggests adding context to the information presented

<Jose> +q

Yolanda: look at the dimensions of provenance and associate the provenance terms with these dimensions

Ivan: RDF descriptions of the mappings is very useful.
... important to have the bridge from one vocabulary to the other

Olaf: understands the usefulness of the exercise -- may be redefining the meaning of the terms in the original vocabs.

Ivan: may be consult the editors of the original vocabs.

<pgroth_> we're over time :-)

Olaf: could do this for models and vocabs of editors who participate in our group

Jose: experiment to highlight the black spots in vocabs to defining provenance.

Yolanda:

trackbot, end telcon

- DRAFT -

Provenance Incubator Group Teleconference

28 May 2010

Attendees

Contents

Summary of Action Items

Scribe.perl diagnostic output