Voice and language recognition yields city information

New in town and don’t know a soul! An evening free, but no idea where to go. What do you do? According to CATCH-2004, you consult one of their interactive systems for the information you need in your native language, and go straight to your preferred venue!

Working prototypes in Athens, Cologne and Helsinki

The CATCH 2004 IST project aimed to provide EU citizens with interactive, multilingual and multimode access to a wide range of information services and systems, offered by providers from both the public and the private sector.

By project close in June 2002, the participants had developed working prototype systems in three European cities, Helsinki, Athens and Cologne. These systems offered information on city events such as sporting activities, local cultural happenings and even TV and radio programme data, and could all be interrogated and reply in more than one language.

Participants in the project included IBM France (lead contractor), additional IBM technical units in Athens, Heidelberg and Prague, Nokia of Finland and OTE in Greece, as well as the Olympic Committee and the cities of Athens, Cologne and Helsinki.

Voice and language recognition

To achieve their ends, the project participants developed a multilingual interface that could handle both direct and spoken-language interrogation, a unified architecture that could handle input from a variety of client devices, and voice-enabled access to the Web databases connected to this architecture.

Key to the success of the system were the voice and language recognition capabilities, based on IBM’s proprietary ViaVoice™ technology, that were inbuilt. This core technology allowed the system to recognise a user’s native language without the need to select from any menu. The system was able to recognise the user’s language simply from the first sentences spoken, then would switch automatically to that language.

In addition, the CATCH-2004 system was a multimode interface, in that this information could be provided not only via city information kiosks, but also via the telephone, using either landline or mobile phones, or the PC. Interaction with the user had to be possible via speech, text or graphics, depending on the characteristics of the access device.

“It was the multilingual and multimode aspects together that were the real test,” says project coordinator Jacques Saint-Blancat of IBM France. “Integrating these two capabilities together into one system, and then making sure it all worked for the system trials, was quite a challenge.”

Multilingual information interface

In Helsinki, the system was set up as a ‘City Events Information’ source that could be interrogated either in Finnish or in English. The application proved so successful that it was extended to include details of radio and TV programmes, allowing users to consult radio and TV programme information on various channels, and enabling them to ask for a list of programmes based on programme name, type, channel, date, time and performer.

The Athens installation was provided with two distinct applications, one focusing on local cultural activities, the other on Olympic sporting events in preparation for the Athens Olympics in 2004. Here the project participants were provided with a database of information on events at the Olympics, for which they developed a voice XML interface based on the contents. Both applications could be consulted using either a Web browser or the telephone. Equally, both applications were capable of handling queries in three languages, English, German or Greek.

Results incorporated into working systems

During system trials, the Helsinki installation handled around 300 calls during the month of October 2001 alone. The results from this pilot were then analysed and used to fine-tune the application, and the final version was exhibited at the IST 2001 conference and exhibition. In addition one of the participants, the city of Cologne, has integrated parts of the CATCH-2004 technology into its existing city information booths.

The project has certainly helped the participants involved, believes Saint-Blancat. “The experience has been very positive for IBM. For our part, what we have learnt has helped us give our voice-response systems greater reliability, as well as making them more versatile.”
Contact:
Jacques Saint-Blancat
IBM France
Le Plan du Bois
F-06610 La Gaude
France
Tel. +33-492-114719
Email: jsb@fr.ibm.com