Speech technology company in court (2)

Translation software has been a white lie from the start of computing in the forties. Various principles of translation have been tried. Systran was one of the oldest machine translation companies. It had its roots in the Georgetown machine translation effort. The company was set up in La Jolla (Cal.), to work on Russian into English for the United States Air Force in the middle of the Cold War. It worked basically from the grammar principles. Systran was adopted by the United States Department of Defense and the European Commission. But the machine translation effort was heavily reduced after that in 1966 a US Government Committee, ALPAC Report was critical about the progress and advised to reduce the efforts. The company was sold in 1986 to the Gachot family, based in Paris, France. Presently SYSTRAN provides the technology for Yahoo!, AltaVista's (Babel Fish) and Google's online translation services, among others and offers commercial versions of SYSTRAN under the usual operation systems.

In the eighties another approach to machine translation was taken in The Netherlands. The BSO Company started a machine translation department which worked on the principle of an intermediary language, codenamed Distributed Language Translation (DTL). The hope was that by having an artificial language in between the target language and the object language a proper translation could be made. It did not work and in the nineties the project was cut.

The speech technology had made its first impression with the introduction of the Reading Machine for The Blind. It was designed by Ray Kurzweil, who had developed the Optical Character Recognition (OCR), a method of scanning text from paper. For his reading machine he combined the OCR method with a speech synthesis machine. A later version of the speech synthesis machine will be bought by L&H.

In the eighties the speech technology was on the rise in academic circles worldwide. In the Netherlands the Nijmegen University, in Belgium the RUG university and in Germany the Saarbrucken University were hotspots for the development of speech technology. Of course, in Europe languages are important as people in Europe speak many different languages.

Jo Lernout (left) and Paul Hauspie (right), the founders of the speech technology and machine translation company L&H were visionaries, but no computer experts, who could develop compression software and algorithms. So they had to lure scientists to their company, which they started in 1987. And they did. They were able to convince Bert van Coillie, Herve Bourlard and Georges Zanelatto. Bert Van Coillie was developing a pc-voice at the RUG in Gent. The company paid a license fee for the Flemish speech synthesis module. The module became the core of the text-to-speech technology. At the same university the Development Environment for Pronunciation Expert System (DEPES) was developed, an innovative system of sound analysis with which algorithms could be developed in other languages than Dutch and Flemish. DEPES became the secret formula of the company. The speech recognition expert Herve Bourlard transferred from Philips to the start-up. And last but not least, Georges Zaneletto joined the company; an expert in signal processing and (de)compression.

Once the company had a scientific base to work from, it was ready for business. It acquired its competitor Dragon Systems. A few weeks earlier it had bought the Dictaphone Corporation, the leader of the medical transcript market. Dictaphone opened the way to vertical markets of intelligent content management and audio mining. And just before the crash the company had the intention to pick up Interactive Systems Inc, a natural speech recognition company.

Blog Posting Number: 762

Tags: speech technology, translation software