Hi, Georg,
Yes, the likelihood of the model generating the training data is maximized when selecting the best question from many questions. Minimum Description Length (MDL) is the optimization criterion deciding when to stop the top-down splitting procedure. Two useful articles are:
1. Odell, J. J. (1995). The Use of Context in Large Vocabulary Speech Recognition. PhD thesis, Cambridge University.
2. Shinoda, K. and T. Watanabe (2000). "MDL-based context-dependent subword modeling for speech recognition." Acoustical Science and Technology 21(2): 79-86.
Yang Wang
Hi
According to which optimum criteria in the data are the HTS questions selected during decision tree building? Is it some kind of information content or likelihood?
Thanks
Georg
--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.
This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.
Please consider the environment before printing this email.