#iks-project

http://www.interactive-knowledge.org/...

IKS mantra and current status

• Prove the value of the IKS stack
  
  - No value yet identified
• Interact with knowledge on user level
  
  - Some examples made, possibly a demo
• Raise tech capabilities of EU CMS providers
  
  - JCR awareness is a starting point
• Raise the value of interactive knowledge at perception of the user
  
  - Friday requirements workshop may give something
• Provide useful methology for developing and managing a semantic CMS
  
  - Not doable yet

Work packages

• WP1: Benchmarking
  
  • Create a list of "challenges" to CMS providers
  - CMS providers should solve the "challenges" with their systems
• WP2: Use cases
  
  - Find requirements from use cases
• WP3: Research into Requirements
  
  - Deliverable: set of requirements described for software engineers (not semantic web researchers!)
• WP4: Community building

Publicity

• Everything is public unless there is a really good reason for it to be private
  
  - Issue: If you publish information on the web, you cannot later publish it as "original research" in the scientific field
• Use tag "iks-project" with different things, #iks-project on Twitter and Qaiku
• Should the IKS website contents be made public?
  
  - To be discussed by the executive board

Where IKS could provide value

• Managing temporally indexed knowledge objects
• Automatic image tagging
• Named Entitity recognition (see also OpenCalais)
• Semi-automatic tagging of content
• Semi-automatic extraction of available metadata (semantic lifting)
• Semantics in content repositories
• Interacting with real things
• Best practices on publishing Microformats and RDFa from CMSs (as search engines now support them)

Some discussion with @bdelacretaz about how CMSs could handle Microformats. I see two ways:

• Generated from structured data on template level (like Midgard has done since 2005)
• Added using a rich content editor ("Insert person", just like "Insert image" is often now)

Semantic CMS benchmarking (Tobias Kowatsch)

• To be used to evaluate the different CMSs participating in IKS in relation to their semantic capabilities
  
  • Explicitness of semantics of the CMS
  • Quality of CMS design and implementation
  - Utility of the CMS
• Objectives
  
  • Identify current semantic capabilities of CMSs ("state of the art")
  • Test if better semantic capabilities will result in better business value
  - Identify design guidelines for the IKS stack
• Semantic metrics
  
  • Low semantic expression (from low to high)
  • Raw signals
  • Data elements
  • Data objects
  • Syntactical metadata
  • Domain schemas
  • Taxonomies
  • Database models
  • High semantic expression (from low to high)
  • Domain ontologies
  • Upper ontologies
  • Context
  - Semantic web
• Quality metrics
  
  • Functionality
  • Coverage of functional requirements
  • Adequate access control
  • Protection from data corruption
  • Maintainability
  • Code complexity
  • Modularity
  • Documentation (class documentation, change logs) covers relevant information
  • Easy extensibility
  • Portability
  • Adaptability of data structure changes
  • Adaptability to HW changes
  - Adaptability to system software environment changes
• Business metrics
  
  • perceived utility of the CMSs
  • usefulness
  • ease of use
  - intention to use / purchase
• Experimental tasks
  
  • "Define the content you want to create: products with a title, image and price"
  • There will be detailed task descriptions
  • CMS vendors will have to implement and document
  • Semantic experts check the semanticity of the implementation
  • Software engineering experts check the quality of the implementation
  • End users check the utility and business value of the implementation
  • IKS will then produce a matrix of scores achieved by the systems
  • The same tests can be used later to check progress
  • There will be analysis whether semantic or software engineering score affects business score
  - In the end this will result in IKS design guidelines

Benchmark exercises (Benjamin Nagel)

• Design
  
  • Task descriptions (business, semantic and software engineering perspective)
  • Industrial partners will comment
  - Template for describing the results
• Performing the exercise
  
  • Industrial partner will do the exercise
  - HSG, SRFG and UPB will provide mentors
• Schedule
  
  • Task briefs end of May 09
  • Task template revisions and feedback in June 09
  • Final version of task templates, and drafts of actual tasks July 09
  • Final version of tasks August 09, benchmarks may start
  • Status meeting September 09
  - Exercises completed October 09

Discussion:

CMSs already provide semantic information, they just do not follow academic models related to them, and do not expose them in a standard way.

But there is a way forward... IKS can propose what standards, like RDFa, should be used and where.

"The best CMS could stay completely within its own platform of Java or PHP, and still follow guidelines like for this issue we'd recommend using OWL"

Any additional software pushed by IKS increases complexity of the CMSs.

A CMS could get 0 on "semantic benchmarks" because it doesn't for instance use RDF, but still get full score for achieving the ten tasks from IKS benchmarks. Then it'd be still a highly semantic CMS even if it doesn't follow W3C recommendations.

photo from assembly

@bdelacretaz: currently, there is no standard on semantic CMSs, so IKS has to invent stuff

Probably no CMSs currently use semantic technologies fully yet. But maybe through IKS they will

Example task: "Use your CMS to create a product list that contains 10 products of your choice. The product list is based on category, and products are selected for the category"

"Make your content searchable on the web (e.g., the dishes in your menu card)"

"Make your content searchable by specific features (e.g., ingredient in the dish)"

-> These are very basic CMS features. But scores come from how well they are implemented, i.e. content variants (desktop browser, mobile, machine-readable), microformatting etc

The benchmarks will also use a standard design template provided by IKS.

http://www.interactive-knowledge.org/...

Attention profiling and recommendation mining mentioned. We already do some of that with Midgard

Content should be adapted to the context: if user accesses the content from his home there should be a "live chat" feature to enable getting in touch with a product specialist. If he accesses it from the shop, then this is not necessary as there are regular salespeople available.

In Midgard, I'd use context injectors and change templates on the fly to fit the situation.