Visitors: Yoshiaki Ito (TRC Information Integration Interactive System Laboratory) , Ryuichi Oka (TRC Theory and Novel Functions Department) Period: September 11 - 23, 1995 (13 days) Places (Attended conferences): SICS (Swedish Institute of Computer Science), University of Erlangen, EUROSPEECH '95
We made a business trip to Europe for thirteen days from September 11th to 23rd, 1995 for the purpose of presenting the papers and surveying the research trends on speech recognition at the international conference, EUROSPEECH '95, and investigating the current research in the institutes and universities which are the members of RWC program. The following is our report.
Visit to SICS
SICS (Swedish Institute of Computer Science) is a Swedish National research organization and is in the town named Kista located to the north of Stockholm. Kista had many research facilities of electrical manufacturers and other industries, and we felt the town was like a condensed version of Tsukuba research and academic city. In SICS, roughly 70 people including the staff members are engaged in research activities. This time we called on Dr. Nilsson who leads RWC distributed lab at SICS. He explained the outline of research at SICS, and we had a lively discussion on the research concerning the Snake robot, which Dr. Nilsson was now developing. We actually took it in our hand and tried to move the joint part of the robot. It was very interesting experience for us.
Visit to University of Erlangen
The University of Erlangen is in an academic town located to the north of Nurnberg, a town in the south east of Germany. We visited Professor Niemann's lab, which was engaging speech-related research. His lab was involved in the research on speech in general, but the major research themes were related to speech rhythm, that is, prosody, and the dialogue models. The two papers presented in the EUROSPEECH '95 by this lab were also associated with the prosody. We have exchanged our views with Dr. Noth who is a research fellow in charge with regard to a variety of research being studied at the professor Niemann's lab. After that, we observed the demonstrations on speech dialogue using telephones and tracking of mobile objects through the image processing. Finally, we introduced the latest research activities at RWCP to them and had an animated discussion.
Participation in EUROSPEECH '95
The EUROSPEECH is an international conference on the speech sponsored by the ESCA (European Speech Communication Association). It is one of the largest international conferences in the speech-related issues, and is held once in two years in Europe as its name says. This year's conference was the fourth and held in Madrid, Spain, and there were six parallel sessions with 535 research paper presentations during four days beginning on September 18th.
In the poster session for the "Word Spotting," one of the final sessions in the final day, we made a presentation titled "Speech Understanding and Speech Retrieval for TV News by Connected Word Spotting." We repeatedly explained the characteristics and the processing method of our research for 75 minutes using posters. The presentation attracted attention with respect to the usage of actual TV broadcasting in particular.
The impression we had throughout the conference was that the research regarding to the dialogue systems and its applications seemed to have increased this time compared to the former EUROSPEECH '93, which had many research themes related to the recognition methods through the Hidden Markov Model (HMM) . In addition, we came across with research in the new ranges similar to the research for the topic border extraction and the application technology in the signal level, which are now progressed in the RWC Information Integration Interactive System lab. We felt that the "Topic Spotting" and the application of speech processing without vocabulary limitation will may become new research areas in the future.