Helping computers identify real meaning

A major obstacle stands astride the road to advanced internet search and services. To a computer, words or pictures on a page are just the 1s and 0s of binary language. The meaning, context, purpose and authority of the words are completely lost in the binary field. Computers cannot thresh the semantic, or meaningful, wheat from all the binary chaff.

“Right now there are many initiatives to develop a 'semantic' web, a web where computers understand the meaning of information contained on a web page,” says Wernher Behrendt of Salzburg Research and coordinator of METOKIS, a project that investigated use of semantic-web technologies for knowledge-intensive fields such as news services, education and clinical studies.

However, he says, “many of the initiatives overlap each other, or are not compatible with each other, which is slowing down the spread of effective semantic search and services.” The METOKIS solution was to develop a new approach based on Knowledge Content Objects (KCOs) and an open-exchange platform called the Knowledge Content Carrier Architecture (KCCA).

KCOs are discrete units of information, typically representing a web page or even a complex multimedia presentation. The METOKIS model identifies information such as the ownership of the data, the data's purpose, licensing information and details about individual information elements within the page. The KCCA provides the infrastructure that allows KCOs to find, identify and interact with each other, combining to provide new services.

“We developed KCOs from an existing ‘foundational’ ontology called DOLCE, which sets out all the elements relevant to an individual Knowledge Content Object. Users can pick and choose whatever elements are relevant to their purpose,” says Behrendt. “In some ways, the KCO is comparable to books – what you put in between the covers is up to the author, but once it is in the shape of a book, you can store it, sell it, look up what it is about, check whether it belongs to you, and of course, read it! The good thing about KCOs is that they enable the computer to do the same!”

Over time, KCOs and KCCA-like systems could provide a method for finding and using any information or any service automatically. Behrendt gives the example of education publishing. The METOKIS partners developed a working demonstrator for educational publishing, one which could allow a school to automatically identify and purchase information from various content sellers for the curriculum of a particular course. All without human intervention.

The project partners also constructed demonstrators for news services and the development and analysis of clinical trials, a particularly challenging area. “In clinical trials, no two studies are alike, even if they deal with the same drug and the same disease. It makes it very difficult to collect clinical trials together to identify broad trends,” says Behrendt.

A KCO-based system would enable researchers to compare studies much more quickly, helping them overcome the difficulties of synthesising the results from several clinical trials into a 'meta-study'. Such meta-studies can be enormously valuable, since they draw information from much larger samples than individual studies.

One side effect of the project work was the development of a clinical trials ‘wiki’ (a website where visitors can add or edit content). This new kind of ‘semantic wiki’ helps researchers to design clinical studies, establishing the number of patients required, the breakdown in their ages, gender, condition, and frequency of therapy.

Another key METOKIS innovation was the development of a methodology to analyse the business argument for new semantic services, something which is an essential but often neglected element of technology research. “We developed tools and methods that can help businesses to establish if a new semantic service provides extra value,” says Behrendt.

The METOKIS project was extremely ambitious, essentially creating a new model for semantic services. “Our reviewers were impressed with the work we achieved, they said the scope of our project could have taken thirty six months, whereas the project only ran for twenty two months,” says Behrendt. “As a result, we are now writing the papers we did not have time for while the project was still running.”

Even so, participants in other projects are already excited by the KCO model and intend to incorporate it into their own research. QVIZ is a cultural heritage project developing a semantic framework for archival services, and it will use the KCO model, as will the Integrated Project LIVE in the broadcasting sector. Similarly the aceMedia integrated project for mobile content will also base its content objects on the DOLCE model.

The success of METOKIS has also prompted the launch of a new two-year project in Austria. The GRISINO project combines GRID computing, semantic web services and intelligent objects based on the KCO model. Here the KCCA infrastructure will be re-implemented using a brand-new Semantic Web Services Model developed within another integrated project called DIP.

“The strategic partnerships with big research and industry players are beginning to pay off in terms of impact – a large Spanish telecommunications company wanted to join METOKIS after the project had started, and they have now joined QVIZ in order to take part in the KCO idea.” says Behrendt.

Source: Based on information from METOKIS