An article in the WSJ has triggered concern over voice strain or "vocal RSI" when using speech interfaces to computers. Long periods of talking softly or in a low vocal range are said to cause strain. It's best to talk in a normal voice. See Pascarelli and Quilter's "Repetitive Strain Injury: A Computer User's Guide" (Wiley, 1994). [Brad Hurley , comp.speech, 8/28/95.] Ron W. Channell of Brigham Young's speech pathology department sees no cause for concern. Speaking, singing, or yelling at an extremely high or low pitch or increased loudness can lead to vocal fatigue, protective/scar tissue growths (nodules), tearing, or bleeding. A soft voice does not require increased tension and is an effective vocal rehabilitation therapy. Don't whisper; just talk as though you don't want to be overheard. An intentional increase in breathiness will also reduce vocal tension. [, 8/30.]

The Linguistic Data Consortium (LDC) and NIST have released a new Air Travel Information System dataset, ATIS3 TEST DATA, under the ARPA Spoken Language Systems (ARPA-SLS) technology development program. or FTP info from pub/ldc on ftp.cis.upenn.edu. [LDC Office , NL-KR, 8/16/95.] (See our RSW digest last week for the full announcement.)

The UWales Speech and Image Processing Research Group is about to collect an audio-visual database of talking subjects, for research in automatic lip-reading, multimodal speech or person recognition, multimodal data compression, and facial motion analysis. They would like to hear about researchers' needs, and about any existing archives. Contact Claude C. Chibelushi , +44 (0)1792 205678 x4698, +44 (0)1792 295686 Fax. [comp.multimedia, 8/28/95.]