Rich Transcription Spring'05 Evaluation

^{June 15, 2005}
Members of ICSI's Speech Group recently participated in the Rich Transcription Spring'05 evaluation, RT-05S, sponsored by the National Institute of Standards and Technology (NIST). This annual evaluation targets the automatic processing of audio recordings from meetings, establishing benchmarks in core recognition capabilities. ICSI submitted systems for both the speech transcription task (or "Speech-to-text") and the speaker segmentation and tracking task ("Speaker Diarization"), and peformed very well in both categories. The creation of these systems was a team effort in collaboration with SRI International. Official results will be made publicly available by NIST at the upcoming MLMI'05 (Machine Learning and Multimodal Interaction) conference in Edinburgh in July.