Журнал «Современная Наука»

Russian (CIS)English (United Kingdom)
 MOSCOW +7(495)-142-86-81

Use recurrent neural networks for ranking list hypotheses in speech recognition system

Kudinov Mikhail   ( graduate student of the Federal Research Center of RAS IU)

The article presents the preliminary results of the use of recurrent neural networks for language modeling on Russian material. It solves the problem of ranking equally recognition hypotheses. To reduce the sparsity of data models were assessed for lemmatizovannom news package. It is also used to predict the morphological information. For the final sorting was used support vector for ranking. The article shows that the combination of neural networks and morphological model gives better results than a 5-gram model with smoothing Knessera-Ney.

Keywords:language model, a recurrent neural network, inflected languages, ranking the hypotheses, speech recognition
 

Read the full article …

Citation link:
Kudinov M. Use recurrent neural networks for ranking list hypotheses in speech recognition system // Современная наука: актуальные проблемы теории и практики. Серия: Естественные и Технические Науки. -2016. -№02. -С. 52-57
LEGAL INFORMATION:
Reproduction of materials is permitted only for non-commercial purposes with reference to the original publication. Protected by the laws of the Russian Federation. Any violations of the law are prosecuted.
© ООО "Научные технологии"

Terms Publications

Issues of the journals (Archive)

ECONOMICS AND LAW
HUMANITIES
NATURAL AND TECHNICAL SCIENCES
COGNITION

Серия - Natural and Technical Sciences