WWW 2004 trip report

Interaction Design and the Semantic Web Workshop

Discussion: There was quite a lot of interaction and discussions during the workshop.
Some interesting points are:

With regard to exploring the semantic space the crucial question is how a query should look like? It was noticed that we should abstract from the traditional keyword search and try to find an approch where possibilities offered by semantic annotations will be fully exploited. The underlying motivation is: why to create all these semantic annotations if we are still using a keyword search?
How annotations can become explicit? The work in this direction is done already as shown by (examples - discussed later):
- Both prototypes shown by Ben and James;
- Poster paper "Multi-Layer Active Documents for the Semantic Web"
- Vannotea tool
- SampLe
- w3photo project
The idea to have as retrieval results concepts rather than documents/images etc. It can simplify the process of query formulation and refinement.
FOAF project was demonstrated. The Friend of a Friend (FOAF) project is about creating a Web of machine-readable homepages describing people, the links between them and the things they create and do.
The problem of interaction with web services. One possible solution presented in the workshop paper "Semantic Web Inreraction on Internet Relay Chat" by Jennifer Golbeck and Paul Mutton. The IRC bot as a mechanism for human interaction with the Semantic Web – specifically with web services and knowledge bases – is simple to program, has an intuitive, conversational interface for human users, and fits well with the inputs and outputs of Semantic Web queries.
It was noticed that an approach for presenting a semantic space which uses lists together with visualizations is extremely useful. (For example ClusterMap visualization techniques).
The use of semantic annotations to create patterns for existing semantic content with the purposes of re-use is discussed in the workshop paper "WrapperInduction for EndUser Semantic Content Development" by Andrew Hogue, David Karger.

Duscussion lead by Web Ontologist, James Hendler and HCI Guru, Ben Shneiderman

The main distinction in their vision about design and annotations was that Ben votes for developing systems
for specific domains. In this way tasks specific for a particular domain can be taken into account.
This will result in better annotations and thus better ontology-based mechanisms can be developed. James
on the other hand has stress that nowadays it is the common practice in research that everybody builds
his own ontologies, annotation schemas, domain-specific systems that does not lead to scalability and
interoperabilty.In his opinion not enough attention is drawn towards developing mechanism and annotations
for cross-domain and cross-community interoperability. (An example was a portal for a group of people
with different research interests). They both presented the prototypes of systems for annotating
photos which looked very similar. PS: do not use the term "end user" with Ben! Posters:

"Multi-Layer Active Documents for the Semantic Web"(poster version / extended version)
Vitaveska Lanfranchi, Fabio Ciravegna, Daniela Petrelli

The paper describes ideas very similar to SampLe. Their framework (documen drafting / writing / publishing / maintaining) can be easily mapped to the SampLe framework with the difference that we apply our ideas to multimedia presentatins and they work with active documents. They also explore the idea of having different annotation layers representing different levels of details. Each layer implies a corresponding visualization. With regard to that they differentiate expext/novice users. The first author will be giving a talk in our reading club scheduled for June 18.

"Is a Picture Worth a Thousand Clicks? Challenges of Adding Semantic Data to Images"
Greg Elin

Presents w3photo.org project. It was initiated with the idea to add semantic annotation to conference photo collections. The environment allows selection of different photo fragments for annotations. The paper correlates with the work done within DSTC on FilmEd project. Ronald Schroeter et. al. have developed a Vannotea tool (is described in Ronald's master thesis "Collaborative Video Indexing, Annotation and Discussion over High-Bandwidth Networks" - is available on my desk) Unless this research involves video, at the bottom level annotations are assigned to shots and fragments of shots which is similar with image annotations.

The Conference

Session: Usability and Accessibility

There were really high quality presentations in this session

Paper: "SmartBack: Supporting Users in Back Navigation" by N. Milic-Frayling from Microsoft research et.al.

SmartBack button complements the standard back button and enables to jump directly to the key pages. Key pages are determined based on structure of a user's navigation trails. The paper presents backgroud studies on navigation paths and discusses implementation decision. The evaluation has shown that back button was used 3 times more often than the SmartBack button (424 vs. 129 clicks).
Overall finding:

The SmartBack button was found useful in "hub and spoke" navigation
Users sometimes had problems understanding how hubs were created
Users who found the SmartBack button functionality not matching their expectations stopped using it

Other two papers from the session discuss various accessibility issues to make web content more easily availlable and adaptable for people with disabilities, elderly people or highly mobile individuals. The second paper presents interesting approach for creating audio browsable content.

Paper: "Web Accessibility: A Broader View" by John T.Richards and Vicki L.Hadson from IBM T.J. Watson Research Center, NY

Unless the paper has very "high-level" title, it discusses concrete issues about how to address common disabilities with regard to interaction problems. Transformation and adaptation are applied to:
- font sizes and families
- spoken text
- lines and letters spacing
- colors and layouts
- keyboard (one hand, typing adjustments etc.)
- page linearization
The idea is that it should be easy for people with no or few experience with computers to make all necessary adjustments to the content appearance. For example changing a text size for the whole page can require a number of operations in a usual web browser. Their browser's Settings button brings up a panel where different text sizes are visually represented which makes a selection process easier. The evaluation has shown that this approach improved legibility and was particularly appreciated by people with developmental disabilities. This paper can be opposed to papers describing user adaptation issues, where complicated solution are proposed. First it stesses the point that the question of providing general accessibility is currently overlooked. And while a simple solutions can be proposed in this direction we keep building complecated system that adapt to different user groups (where different usually means anything but not people with specific needs for such a basic task as web browsing).

Paper: "HearSay: Enabling Audio Browsing on Hypertext Content" by I.V. Ramakrishnan et.al. Stony Brook University, NY

The paper introduces 2 key technologies: automatic partitioning of Web documents through structural and semantic analysis and usage of recently standardized VoiceXML (W3C Recommendation since March 16th) to represent voice dialogs automatically.

Session: Semantic Web Applications

Paper: "Building a Companion Website in the Semantic Web" by T. Miles-Board, C. Bailey, W. Hall, L. Carr, University of Southampton

The surprisingly disappointing talk by L. Carr where a lot of talking did not shed the light at the essence of work.

The idea of the paper is interesting though:
In contrast with most of the sites that provide just an electronic equivalent of a (text)book, a companion website aims to provide an on-line support for undestanding the (text)book's content and also helps to keep the content of the book up-to-date. They make use of Bloom's Taxonomy (a hierarchy of intellectual skills involving the acquisition and use of knowledge in the cognitive domain) to find out the features the companion website has to offer. The authors are interested in discovering the authoring and presentation techniques that could be employed to create "successful" companion website.
The paper contains an extensive review of companion websites with the analysis what features were or were not implemented by their developers. The review shows that those sites provide very little support for both students and instructors groups. Besides almost non of them take a full advantage of hypertext in order to provide better navigation.
4 techniques for authoring and presentatioin were identified which facilitate the reader's progression to the higher-level educational objectives:
- uodate / extend the text itself;
- integrate the textbook material with other (newer) resources by linking to them;
- use hypertextual devices to expose the underlying organizational structure of the material;
- create exercises which explicitly help readers assess their learning achievements or help instructors to assess students;

The authors argue that together with these techniques the posiibilities offered by the SW should be used such as dynamically synthesize material from different sources, use semantic services to facilitate users' activities etc.

Paper: "CS AKTive Space: Representing Computer Science in the Semantic Web" by m.c. schraefel et.al. University of Southampton

Intertaining presentation by Monica and Nigel where Monica was trying to put their work into the context of the current SW applications and Nigel was "sailing" CS AKTive Space.