Ueberla, J. "Analyzing and Improving Statistical Language Models for Speech Recognition". PhD thesis, Simon Fraser University, Vancouver, Canada, 1994. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Hierarchical Statistical Language Models: Experiments On.. - Galescu, Allen (2000) (Correct)
....(PP) to compare the performance of the two models ( 17] Although we computed PP values for all models, we only give them for the curiosity of the reader. The actual comparisons are made in 1 Compare to the ones given in [9] terms of adjusted perplexity (APP) a measure introduced in [19], which adjusts the value of PP by a quantity dependent on the number of unknown words in the test set, and the number of their occurrences. We compute the APP value on the full test set, and thus we can compare two models with different vocabularies. We again refer the reader to [5] for more ....
Ueberla, J. "Analyzing and Improving Statistical Language Models for Speech Recognition". PhD thesis, Simon Fraser University, Vancouver, Canada, 1994.
Detection and Transcription of OOV Words - Fetter (1998) (3 citations) (Correct)
....W = argmax W P(A j W ) P(W ) 2.4) Equation 2.1 requires a model for all possible acoustic signals A, whereas Equation 2.3 (or 2.4) requires a model that covers every possible word sequence W . The latter is much easier to compute because W is discrete, whereas A is a continuous signal [Ueberla, 1994]. For this reason it is easier to apply Equation 2.3 (or 2.4) to obtain the most probable word string W . Equation 2.3 (or 2.4) also allows a modular strategy for speech recognition. The acoustic information contained in the speech signal is captured by the acoustic models, which provide the ....
....conveyed in the history of a word have also been investigated, and resulted in other language modeling paradigms. Among them are decision trees, caches, and maximum entropy models, to name but a few. For a more thorough discussion of existing language models, see for example [Rosenfeld, 1994] or [Ueberla, 1994]. We will now focus our attention on N gram language models, and in particular, on the estimation of its N gram probabilities. The N gram probabilities, and the conditional probabilities of a word given the N Gamma 1 preceding words can be estimated using the relative frequency of occurrence of ....
Ueberla, J. (1994). Analyzing and Improving Statistical Language Models for Speech Recognition. PhD thesis, Simon Fraser University, Vancouver, British Columbia, Canada.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC