Journal article

DEORAS Anoop, MIKOLOV Tomáš, KOMBRINK Stefan and CHURCH Kenneth. Approximate inference: A sampling based modeling technique to capture complex dependencies in a language model. Speech Communication. Amsterdam: Elsevier Science, 2012, vol. 2012, no. 8, pp. 1-16. ISSN 0167-6393. Available from: http://www.sciencedirect.com/science/article/pii/S0167639312000969#
Publication language:english
Original title:Approximate inference: A sampling based modeling technique to capture complex dependencies in a language model
Title (cs):Přibližná inference: podchycení složitých vztahů v jazykovém modelu pomocí techniky založené na vzorkování.
Pages:1-16
Book:Speech Communication
Place:NL
Year:2012
URL:http://www.sciencedirect.com/science/article/pii/S0167639312000969#
Journal:Speech Communication, Vol. 2012, No. 8, Amsterdam, NL
ISSN:0167-6393
Publisher:Elsevier Science
URL:http://www.fit.vutbr.cz/research/groups/speech/publi/2012/deoras_SpeechCommunication2012_1-s2.0-S0167639312000969-main.pdf [PDF]
Keywords
Long-span language models; Recurrent neural networks; Speech recognition; Decoding
Annotation
This paper deals with approximate inference: a sampling based modeling technique to capture complex dependencies in a language model
Abstract
In this paper, we present strategies to incorporate long context information directly during the first pass decoding and also for the second pass lattice re-scoring in speech recognition systems. Long-span language models that capture complex syntactic and/or semantic information are seldom used in the first pass of large vocabulary continuous speech recognition systems due to the prohibitive increase in the size of the sentence-hypotheses search space. Typically, n-gram language models are used in the first pass to produce N-best lists, which are then re-scored using long-span models. Such a pipeline produces biased first pass output, resulting in sub-optimal performance during re-scoring. In this paper we show that computationally tractable variational approximations of the long-span and complex language models are a better choice than the standard n-gram model for the first pass decoding and also for lattice re-scoring.
BibTeX:
@ARTICLE{
   author = {Anoop Deoras and Tom{\'{a}}{\v{s}} Mikolov and Stefan
	Kombrink and Kenneth Church},
   title = {Approximate inference: A sampling based modeling technique
	to capture complex dependencies in a language model},
   pages = {1--16},
   booktitle = {Speech Communication},
   journal = {Speech Communication},
   volume = {2012},
   number = {8},
   year = {2012},
   publisher = {Elsevier Science},
   ISSN = {0167-6393},
   language = {english},
   url = {http://www.fit.vutbr.cz/research/view_pub.php?id=10160}
}

Your IPv4 address: 54.156.92.243
Switch to IPv6 connection

DNSSEC [dnssec]