Tuning of Acoustic Modeling and Adaptation Technique for a Real Speech Recognition Task

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F19%3A43956404" target="_blank" >RIV/49777513:23520/19:43956404 - isvavai.cz</a>
Result on the web
<a href="https://link.springer.com/chapter/10.1007/978-3-030-31372-2_20#aboutcontent" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-030-31372-2_20#aboutcontent</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1007/978-3-030-31372-2_20" target="_blank" >10.1007/978-3-030-31372-2_20</a>

Result language
angličtina
Original language name
Tuning of Acoustic Modeling and Adaptation Technique for a Real Speech Recognition Task
Original language description
At the beginning, we had started to develop a Czech telephone acoustic model by evaluating various Kaldi recipes. We had a 500-h Czech telephone Switchboard-like corpus. We had selected the Time-Delay Neural Network (TDNN) model variant “d” with the i-vector adaptation as the best performing model on the held-out set from the corpus. The TDNN architecture with an asymmetric time-delay window also fulfilled our real-time application constrain. However, we were wondering why the model totally failed on a real call center task. The main problem was in the i-vector estimation procedure. The training data are split into short utterances. In the recipe, 2-utterance pseudospeakers are made and i-vectors are evaluated for them. However, the real call center utterances are much longer, in order of several minutes or even more. The TDNN model was trained from i-vectors that did not match the test ones. We propose two ways how to normalize statistics used for the i-vector estimation. The test data i-vectors with the normalization are better compatible with the training data i-vectors. In the paper, we also discuss various additional ways of improving the model accuracy on the out-of-domain real task including using LSTM based models.
Czech name
—
Czech description
—

Project
<a href="/en/project/EF16_013%2F0001781" target="_blank" >EF16_013/0001781: LINDAT/CLARIN - Research infrastructure for language technologies - extension of the repository and its computational power</a><br>
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year
2019
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
Statistical Language and Speech Processing, 7th International Conference, SLSP 2019, Ljubljana, Slovenia, October 14–16, 2019, Proceedings
ISBN
978-3-030-31371-5
ISSN
0302-9743
e-ISSN
1611-3349
Number of pages
11
Pages from-to
235-245
Publisher name
Springer
Place of publication
Cham
Event location
Ljubljana, Slovenia
Event date
Oct 14, 2019
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—

Similar results(10)