Haute couture for transcription´s next top model: text processing technologies on the catwalk

Summary form only given. The top model metaphor refers to forward-looking applications expected to come of age in the next few years. They are likely to involve the large-scale automatic transcription of lectures, meetings, podcasts, oral histories, etc. In the realm of podcasting alone, free-form content searching could quickly become critical at the current growth rate of the medium. Such applications entail many speech recognition challenges, which John Hansen comprehensively covers in the first talk of the forum. But they also raise a number of issues which are somewhat orthogonal to the process of transcription itself. This talk, accordingly, serves as a catwalk presentation of the various component technologies involved. It points to recent work in each area, highlights major bottlenecks, covers promising solutions, suggests some new areas of research, and offers some perspectives regarding comparative merits and associated trade-offs