A few weeks ago, about 150 professionals from the Dutch-speaking speech and language technology sector gathered in Antwerp at the second Stevin Gathering organized by the Dutch-Flemish Taalunie (= Language Union). For each group of three visitors from the Netherlands, there were two visitors from Flanders. Roughly 60% came from academia or government bodies, 40% from industry.
The most interesting morning talk was given by keynote speaker Nick Wright from the Edinburgh Stanford link. In his presentation on Translating Language Technology into Business Benefits, mr. Wright pleaded for an evolutionary approach to introduce language technology into the European market environment. The best strategy to bring language technology from the lab to the marketplace is to improve existing products and processes, by making them faster, better and/or cheaper. In short: don’t focus so much on (selling) technology, but rather on its applications and on the concrete business value that these represent. Mr. Wright explicitly mentioned document translation and call center automation as two obvious target domains where [speech and] language technology are making and will continue to make a big business difference. To speed up technology transfer, research and development programs should shift focus from the former to the latter. In mr. Wright’s own words: R&d should become more and more r&D.
After his presentation, mr. Wright invited questions from the audience, but received none. Was it that the industry people present represented almost exclusively the speech and language technology providers, rather than the end users? There indeed were hardly any call center or telco representatives on the attendance list. No big surprise, on the other hand, as the academic institutions are explicitly meant to be the primary beneficiaries of the Stevin program in terms of direct funding.
To be fair, it should be mentioned, however, that w.r.t. technology transfer, the Stevin program committee did not limit itself to inviting a keynote speaker. Stevin will also directly fund the development of a few end-user applications, in order to increase the visibility of speech and language technology in the Dutch-speaking market. More details on this aspect of Stevin’s side policy (flankerend beleid) are to be announced in May 2005.
In the afternoon I attended the workgroup on the Development of a Speech Recognizer for Dutch. Although the pressing need for a speech recognizer for Dutch (as opposed to other languages) is not clear, at least not to me, multiple researchers did plead very strongly for a generic ASR platform that was readily available (in terms of IPR), as well as extensible and configurable (in terms of technology). The audience explicitly asked for domain and application requirements to be kept out of the tender, which seemed to conflict somewhat with the Stevin program committee’s unwillingness to fund another development-only platform. The extent to which the to-be-developed ASR platform will have to be ready for immediate product development, remained an open question after the workgroup. To be continued …