Search

27 Jul 2020: This model uses a long-range history and exploits topic information in the test text to adjust probabilities of test words. ... The PLSA-based model was found to lower test set perplexity over a traditional wordclass-based 4-gram by 13% (optimistic

27 Jul 2020: The focus is primarily on retrieval studies, and on speech tests directly related to retrieval, not on speech recognition itself. ... The report draws on the many and varied tests done during the project, but also presents a new series of results

27 Jul 2020: set. These factors control the point at which training set performance or test set generalisation become the limiting factors. ... The experimental results provide an empirical verification of the training set/test set performance of frame discriminative

27 Jul 2020: all the test data together or 8.6% when processing the test data shows independently. |

27 Jul 2020: The complete system yields an overall word error rate of 22.0% on the 1996 unpartitioned broadcast news development test data and just 15.8% on the 1997 evaluation test set. |

27 Jul 2020: For high dimensional inputs the performance of the single layer network alone is better on the test data, whereas for low dimensional inputs a two layer adaptive network is required and ... this needs much more training before it can produce a better

27 Jul 2020: It consists of speaker-independent (SI) read material, split into training, development test and evaluation test sets.

27 Jul 2020: An initial series of tests are performed in vitro. The statistical theory to predict the increase in signal to noise ratio is developed and verified empirically. ... Having proven the benefits of spatial compounding in vitro, tests with higher levels of

27 Jul 2020: word bigram "hub" tests and the second lowest error rate on the 20k word trigram "hub" test.

27 Jul 2020: word error rate on the 1998 evaluation test set.