AT&T Co plans to exploit rather more of the technology it has developed in the near future, Data Systems president Robert Kavner told Esther Dyson’s Personal Computer Forum in Tucson, Arizona last week and to demonstrate what he had in mind, he showed off some of the company’s new speech-recognition technologies: one demonstration, reports Microbytes Daily, was of a connected-word speech recognition system on an 80386-based AT-alike with microphone and expansion board containing AT&T signal processors, in which an AT&T employee posed as a bond trader presented with a fill-in-the-blanks screen and was able to fill in the data for each field by speaking the data without pausing between words and in any field order; that system is speaker-dependent, but Kavner also showed off a speaker-independent speech recognition system with the same set of hardware plus a telephone – the demonstrator phoned in and an electronic operator asked him to choose between collect, calling card or information, and was able to be understood when he said I would like to make a collect call; the system repeated collect to verify that it had spotted the key word in the sentence – and the system even managed to handle a thick Japanese accent.
