Open Vocabulary Confusion Networks for Speech Recognition
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Christian Steinruecken.
A limitation of most current speech recognizers is that they only recognize words from a fixed vocabulary. If the user utters a novel word, the recognizer can only indicate it doesn’t know, or choose the “least-worst” word in its vocabulary. In this paper, I explore a technique for addressing this deficiency using automatically derived units made up of letters and phones. I show how these units can be merged to form likely new words. I further give an algorithm for doing this merger while maintaining a word lattice structure. This allows construction of a word confusion network containing both in- and out-of-vocabulary (OOV) words. Experiments show these open vocabulary confusion networks provide improved word recognition rates of up to 24% relative.
Further details can be found in the paper or in chapter 6 of my thesis
This talk is part of the Inference Group series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|