Speech Trends and Predictions, or Do We Need Text?

Abstract

There has been remarkable progress in automatic speech recognition technology in the last ten years. Performance, however, remains significantly worse than human performance. Speech recognition systems do reasonably well for careful speakers in relatively quiet environments when the conditions do not vary much from the training conditions. However, degradation can be severe with even minor levels of noise, vocabulary, and other characteristics. Nonetheless, speed and accuracy of the automatic systems have increased to the point where many useful applications are possible. The rapid growth in the number of companies devoted to speech recognition applications attests to this growth in performance. This brief report explores the further potential for speech technology.