Meetings/20070416/DefiningImageAccess-ECS-Southampton

From ImageWeb

Jump to: navigation, search

Contents

Southampton ECS visit

David Shotton, Graham Klyne and Jun Zhao visit to Southampton ECS, 16 April 2007.


SERPENT images ePrints respository

  • Jessie Hey
  • (who else?)

Spent some time digging in Serpent image archives - note new URIs (in wiki now). Spoke to Tim (Brody?) - apparently, the SERPENT team are in the process of cleaning up their metadata and don't intend to set up the OAI-PMH interface until that is done. Subject to the research team's agreement, Tim can provide us with an XML dump of the ePrints data.

ACTION GK/JZ: email Tim to ask for access to raw XML metadata (on the assumption we can get a sense ofg what the OAI might look like). Formulate a request that Tim can present to the project team.

KULTUR - new arts linkup project - led by libraries at Soton. Mark Brown. http://www.jisc.ac.uk/whatwedo/programmes/programme_rep_pres/repositories_sue/kultur.aspx.

Looking for alternative collections with images and varying levels of metadata. Use ROAR to list repositories. PRESERV project (one of Jessie's projects) has profile of file formats in each repository. Uses PRONOM to determine file types(?):

The resulting page is a list of OAI-PMH links and resource links for the corresponding file type.

Questions about ROAR: ask Tim Brody

ePrints handling of multiple files/record - one metadata record can associate with multiple files - no overall wrapping framework is applied.

Coreference in SERPENT: deposition process uses a combination of time, geo location, other metadata.

Chris Gutteridge - Web Projects Manager; looks after ePrints, etc.

Seminar presentation

David and Graham presented about our data webs work. The presentation can be seen here:

Feedback positive. Commentators liked the separation of schema alignment and coreference.


ePrints and domain specific metadata

Corridor discussion with Chris Gutteridge (ePrints lead developer), following the seminar. He raised the issue that serving a specific metadata schema through OAI-PMH requires repository administrator action, so serving arbitrary domain-specific schema creates undesirable management overhead for an institutional repository.

He proposes an alternative approach:

The ePrints software, as I understand do other repository systems, allow an arbitrary metadata schema to be served via OAI-PMH. But to enable serving of any particular metadata schema, ePrints requires configuration action on the part of the repository administrator. A consequence of this is that it is impractical to serve domain-specific metadata via OAI-PMH from an institutional repository housing maybe hundreds or thousands of individual collections, each potentially with their own domain-specific metadata schema.

Chris Gutteridge proposes an alternative approach: that domain-specific metadata might be stored as part of the content, as a separate file, and a couple of specific metadata terms be introduced to indicate: (a) that a given file is a metadata record, and (b) the format and schema of the metadata

This approach seems to be particularly apposite with ePrints, where one repository record (roughly, one digital object) can have a number of associated files, each of which has a URI for retrieval. This is commonly used, for example, to create a record with a collection of images about a specific subject at a specific time and a single set of repository metadata entries for that collection - it's a very simple form of composite object (not at all at the level of compound digital object standards like MPEG-21 DID, but easy to implement, understand and use).

Thus, domain-specific metadata, or indeed any metadata not supported at the repository level, can be packaged alongside the raw data (images, documents, etc.) and distinguished by the presence of just a couple of additional "core" metadata fields.

(see also: http://lists.ontonet.org/mailman/private/bioimage/2007q2/000664.html)


OpenMKS

  • Matthew Addis
  • Patrick Sinclair
  • Paul Lewis

Lessons: processes - OpenMKS deals with data post-integration.

Doesn't use RDF or semantic web data model. Uses web services stack for component integration - very heavyweight.

Discussed:

  • CRM Core
  • Virtuoso
  • Allegro graph

Data management problems: importance of having the right supporting tools. (I think this is in the context of using a relatiopnal database vs an RDF triple store - relational databases have highly developed management and support tools.)

OpenKnowledge - use P2P ideas and ontology alignment; Edinburgh - Dave ?. Lightweigh coordination calculus for agent sharing of information.

HealthAgents - building a distributed decision support for diagnosis & prognosis of brain tumours.

Bricks - IP FP6 - cultural heritage almost P2P. EPOCH - NoE FP6.

LEAF, SCHEMA - thesauri alignment.

FP6 visual search engine.

EU commission - consider research infrastructures. CLARION(?), DARIUS - european wide integration for cultural heritage.

FP6: IST - capacities (downstream - building working infrastructure).

FP7: ICT(core research) - ..

eChase - business models for using cultural content - but not really viable. Tools exist, but not really mature. Useful lessons in personal papers (not published?).

CRM core doesn't solve all the problems - loses some of the richer descriptive capabilities of CIDOC. Core is probably enough.

Know your users ... but be able to support different classes of users from the same basic information stores.


mSpace, Rich Tags and more

  • mc schraefel
  • Max Wilson

mSpace

Dan Smith, mc -- eprints.ecs.soton.ac.uk -- investigating problem of how easy it is to pull together output from various repositories to gain access to a body of eScrience research -- not just authors and papers, but what projects they work on. More than just digital repositories. Infrastructure challenges -- need good locators, something like JESS(?) system (as model)

... possible collaborative work?

mSpace -- is neutral - just needs a schema ("Ontology lite").

cs.aktivespace.org -- older demo -- similar problem domain. Coreferencing; common AKT ontology.

mSpace API for pointing data source at it.

ds@ecs.soton.ac.uk -- cc mc -- ask for access to mSpace 0.6 (next release mSpace 0.8)

Image types: 2D images, video, ... How to present; e.g. thumbnail+info; or a "menagerie" of images. These are ongoing topics for mSpace. mSpace has a tagging/comment/lightweight blogging + RSS feed.

Annotation -- what are best strategies for rapid annotation and evolution to lexicon? Possible collaboration angle?

Extra metadata - e.g. links to project database - Chris (?) discussion.

Think about: mSpace affords additional information about terms in the schema.

mSpace support of visible context for the human user - column approach to browsing is part, but other part is information overview.

"Preview cues" - highlight over audio - hear clips ... see typical/representative images? "Something better than nothing".

ACTION: David send music paper to mc? (Done.)

Rich tags - "great outcome would be to get related work for free". Just enough contextual information to disambiguate (tags).

Use of micro ontologies. Embed rich-tagger into a site. (Look at proposal -- review deliverables.) Try to use blog ping/trackback ideas to track updates to tags. Discussion of Joanna's idea -- mc: microontologies, can a blog help explore and share an ontology? (doingpad - how do ordinary people create structured data and how doi they share it if they don't have ((things to hang it on)) ...)

ACTION: GK/(JZ?) revisit Rich Tags web site.

A kind of ontology (or micro-ontology?) guided link to related information ("things like")

Jun -- asks about how to present SMW labelling information information. JHas problems to pull it out easily.

Re ontology-fixed interfaces: "You better hope the fields you want are there".

"URI modeller" project proposal - URI formation patterns to resolve some problems of URI usage.

"Ghost in the machine" ECS eprints, sweb workshop

Open ID project

ACTION: send link to JISC call to mc ...

Personal tools
Oxford DMP online
MIIDI
Claros