Semantic Technology Conferences Overview

by Mark Jacobson and Ed Stevenson

Really Strategies' consultants attended two Semantic Web related conferences in March. Mark Jacobson and Jeff Wood attended the Semantic Technologies Publishers Workshop, presented by IDEAlliance in conjunction with the BookTech Conference and Expo, in New York on March 9. Lisa Bos and Ed Stevenson attended the Semantic Technologies Conference in San Francisco from March 7 to 10.

Semantic Technologies Publishers Workshop, New York

About 40 people attended the day-long Semantic Technologies Publishers Workshop, which comprised a key note session and three tutorials. Typical attendees were from organizations that have XML in the workflow and are just beginning to investigate semantic technologies.

The key note speaker, David Wood, from the Semantic Web Research Group inside the MIND LAB at University of Maryland Institute for Advanced Computer Studies, talked about increasing adoption of semantic technologies and cited the Nokia portal (which is being released as open source) and Oracle's RDF Network Data Model. Other presenters were Dianne Kennedy (IDEAlliance), Steve Pepper (Ontopia), and Dean Allemang (TopQuadrant, Inc).

The tutorials focused on basic understanding of metadata, taxonomies, and ontologies and the approaches to implementing them—Topic Maps and RDF/OWL. The day concluded with an Industry Roundtable lead by Mark Jacobson.

Semantic Technology Conference, San Francisco

During the same time, Lisa Bos and Ed Stevenson attended the Semantic Technology Conference in San Francisco. The conference focus was broader than the publishing industry and looked at semantic technologies across industries. There was a mix of theoretical (what is possible or coming) and the practical (what is being done now) with a bit of a heavier emphasis on the former.

Dave McComb, the conference chairman and president of Semantic Arts, Inc, presented "Semantics 101." He highlighted several important fundamentals, including:

Semantics = The study of meaning
The concepts of precision and veracity are main focuses in semantics
Taxonomy = Hierarchical classification of things
Ontology = "A specification of a conceptualization" (Tom Gruber); in other words, an ontology specifies the concepts that relate things together

He also discussed six areas in which semantics are being used today: transactional systems, internal integration, knowledge management, decision support, unstructured activity (like natural language parsing), and "environment" (everything outside your organization).

In his presentation, Eric Miller from W3C illustrated how the trend of the web has moved from the file level (FTP) to the text level (html and links) and is now moving to the data level (XML, RDF, OWL, URIs). Within this he discussed the trends to pull data from various sources to discover new meaning and present in different formats (sounds like content reuse).

Doug Lenat of Cycorp has created an artificial intelligence engine, called Cyc, over the past 20 years. Cyc has more than 3 million assertions (such as "a bird has feathers" or "every child has a biological mother") from which it can then do some reasoning. With this reasoning and the ability to search a vast amount of info over the web, Cyc can then supposedly answer questions such as "what large US city is most vulnerable to an anthrax outbreak?" In one example, Lenat discussed how current searches for an image of a "smiling person" would not find a photo of a man with his daughter with the caption "John watches as his daughter take her first steps." However, Cyc, with its inference capability, could reason that a man must be smiling when his daughter tasks her first steps; and therefore would retrieve the image on a search.

Peter Norvig, the director of search quality at Google, also presented. Obviously, Google has indexed a massive amount of information, but it was really interesting to see this called out in the presentation. They've been able to use the volumes of content they've indexed plus searching patterns to improve on their tools over time.

It was interesting to hear Lenat and Norvig speak back to back. In the photo example mentioned in the Cyc presentation, we could guess that Google would argue that it is "good enough" to get the hundreds of other pictures that more explicitly stated someone is smiling than to get that specific photo. Cyc is concerned with getting the right pictures out of all that exist, whereas Google would probably argue that it is adequate to match the query. (Note this last statement is our own interpretation; neither speaker compared the two systems).