5th International Workshop on Knowledge Markup and Semantic Annotation (SemAnnot'2005)

7th November 2005 (Workshop day), Galway, Ireland

5th International Workshop on Knowledge Markup and Semantic Annotation (SemAnnot'2005), co-located at the 4th International Semantic Web Conference (ISWC'2005)
Author: Raphael
CWI participants: Raphael
# participants: 50

Overall impression

The audience was quite large: about 50 persons. The first part of the workshop (morning) was really text based and linguistics oriented, thus it explain the very few notes I took :-) The second part of the workshop (afternoon) dealt with either photo or video annotation, thus much more interesting for the group. A special mention for the few posters, very interesting. The proceedings of the workshop is available in my office.

Agenda

9:30 - Introduction

Introduction of the forthcoming European funded projects:

• X-Media: http://www.iti.gr/db.php/en/projects/X-Media.html, IP project, 48 months from January 1st 2006
• K-Space: http://www.iti.gr/db.php/en/projects/K-Space.html, NoE, 36 months from January 1st 2006

9:45 - Knowledge extraction from WebPages

Semi-automatic annotation of textual documents using an ontology. Machine learning techniques on the text. First, an expert associates some syntactic structure with semantic concepts of an ontology. Use then the DOM tree of each HTML pages for mapping these syntactic structures with semantic concepts. The asumption is that all the pages are made according to the same structure ... for instance because they are dynamically generated from a database.

10:00 - Semantic Web Development with WSDM

Boring and bla bla stuff ... I was sleeping!

10:30 - Towards the definition of guidelines for RDF and Topic Maps interoperability

Related to the work done in the RDFTM Task Force of the W3C SWBPD group. Good explanation of the content of the note written by this TF.

10:45 - Coffee break

11:00 - Ontological Annotation with WordNet

Ontology construction from WordNet: Verb ontology and Noun Ontology developped in OWL. Construct a classification model to predict the WordNet sense and class for each noun and verb.

11:30 - Feature Representation for Cross-Lingual, Cross-Media Semantic Web Application

12:00 - Exploiting User Queries and Web Communities in Semantic Annotation

SQAPS (Semantic Query-based Annotation P2P Sharing system): exploit user queries to search engines in semantic annotation of web resources. Idea = associate resources to annotated query. Use Wikipedia as a semantic source.

12:30 - Lunch break

14:00 - Using a Multimedia Ontology Infrastructure for Semantic Annotation of Multimedia Content

14:30 - Semantic Integration and Retrieval of Multimedia Metadata

Motivation: The amount of digital media that has to be actually managed has already become unaffordable without fine-grained computerised support. MPEG-7 is the greatest metadata framework created to date but it is based on XML Schemas. Therefore, its does not have formal semantics, which makes difficult to manage, extend and integrate it. Consequently, there have been a lot attempts to move MPEG-7 to the Semantic Web.

Approach: The approach presented here contributes a complete and automatic mapping of the whole MPEG-7 standard to OWL. It is based on a generic XML Schema to OWL mapping. The previous mapping is complemented with a XML metadata instances to RDF mapping that completes a transparent transfer of metadata from the XML to the Semantic Web domain. The generated MPEG-7 OWL ontology is then used as an “upper-ontology” for multimedia metadata, and three different music schemas have then been linked. Thus, it is possible to retrieve related information from instances of all the metadata sources. Furthermore, detecting and merging instances from different sources has allowed to enhance the description of audio files, both content-based and editorial data.

Discussion: I have asked questions about the modeling choices that have to be done while translating from the MPEG-7 XML Schema files to an OWL model. Since the two languages have obviously not the same expressivity, arbritary choice has to be made for performing this conversion. For example, some knowledge, naturally captured by XML Schema such as the ordering and the nesting of particular descriptors, is lost while switching to the OWL space. It was finally a preliminary discussion before the EWIMT workshop.

14:45 - A semi-automated Framework for Supporting Semantic Image Annotation

15:00 - Coffee break

15:15 - Posters

16:00 - A Flexible Approach for Managing Digital Images on the Semantic Web

Presentation of the PhotoStuff tool. Its main functionnalities are:

• ontology-based annotations of images, either on the Web or on your local machine;
• the assertions can be made about the image, its content, or the content of some specific regions in the image;
• the annotations can concern any domain and can be published on the Web.

Discussion: The main criticism I do is the way they represent the localization of a specific region in an image. Actually, they embed a piece of SVG as the value of a property inside the description. I think it is then impossible to refer to this region in another annotation, except to draw again exactly the same rectangle !

16:30 - Time to evaluate: Targeting Annotation Tools

17:00 - Discussion and Wrap up

Poster Presentations

The poster of Jean-Philippe (from INA) was quite interesting. What he did briefly is, for a given broadcast channel:

• gather on the web the scheduled programs grid (from an electronic program guide);
• take the media stream, and analyze it, in order to automatically decompose temporally the stream into its various programs;
• the programs grid is used for such a purpose but it is itself also modified (mainly re-calibrate) in the process;