Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi

Translated title of the contribution: Using hybrid lexicon units and incorporating language constraints in speech recognition

Osman Büyük, Hakan Erdoǧan, Kemal Oflazer

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

We face problems in large vocabulary continuous speech recognition for agglutinative languages, due to lack of coverage of all possible words. Since it is not enough to have a finite full-word lexicon, we may use sub-word units for recognition. Using sub-word lexicon units and developing language models based on these units solves the coverage problem. However, this results in increased acoustic confusability and shorter effective language model history length. We introduce new ways to choose lexicon units and we incorporate linguistic constraints into a statistical language model developed with the new units. We represent both the statistical language model and linguistic constraints as weighted finite state machines (WFSM) and combine them to obtain a novel language model. We study the performance of the new language model and show that it achieves 3% relative reduction in word error rate when used in recognizing a test-set of 2151 words.

Original languageUndefined/Unknown
Title of host publicationProceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005
Pages111-114
Number of pages4
Volume2005
DOIs
Publication statusPublished - 2005
Externally publishedYes
EventIEEE 13th Signal Processing and Communications Applications Conference, SIU 2005 - Kayseri, Turkey
Duration: 16 May 200518 May 2005

Other

Other
CountryTurkey
CityKayseri
Period16/5/0518/5/05

Fingerprint

Speech recognition
Linguistics
Continuous speech recognition
Finite automata
Acoustics

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Büyük, O., Erdoǧan, H., & Oflazer, K. (2005). Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi. In Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005 (Vol. 2005, pp. 111-114). [1567632] https://doi.org/10.1109/SIU.2005.1567632

Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi. / Büyük, Osman; Erdoǧan, Hakan; Oflazer, Kemal.

Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005. Vol. 2005 2005. p. 111-114 1567632.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Büyük, O, Erdoǧan, H & Oflazer, K 2005, Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi. in Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005. vol. 2005, 1567632, pp. 111-114, Kayseri, Turkey, 16/5/05. https://doi.org/10.1109/SIU.2005.1567632
Büyük O, Erdoǧan H, Oflazer K. Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi. In Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005. Vol. 2005. 2005. p. 111-114. 1567632 https://doi.org/10.1109/SIU.2005.1567632
Büyük, Osman ; Erdoǧan, Hakan ; Oflazer, Kemal. / Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi. Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005. Vol. 2005 2005. pp. 111-114
@inproceedings{56a951810061462fb60209fbd6883ed2,
title = "Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin ger{\cc}eklenmesi",
abstract = "We face problems in large vocabulary continuous speech recognition for agglutinative languages, due to lack of coverage of all possible words. Since it is not enough to have a finite full-word lexicon, we may use sub-word units for recognition. Using sub-word lexicon units and developing language models based on these units solves the coverage problem. However, this results in increased acoustic confusability and shorter effective language model history length. We introduce new ways to choose lexicon units and we incorporate linguistic constraints into a statistical language model developed with the new units. We represent both the statistical language model and linguistic constraints as weighted finite state machines (WFSM) and combine them to obtain a novel language model. We study the performance of the new language model and show that it achieves 3{\%} relative reduction in word error rate when used in recognizing a test-set of 2151 words.",
author = "Osman B{\"u}y{\"u}k and Hakan Erdoǧan and Kemal Oflazer",
year = "2005",
doi = "10.1109/SIU.2005.1567632",
language = "Undefined/Unknown",
isbn = "0780392396",
volume = "2005",
pages = "111--114",
booktitle = "Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005",

}

TY - GEN

T1 - Konuşma tanimada karma dil birimleri kullanimi ve dil kisitlarinin gerçeklenmesi

AU - Büyük, Osman

AU - Erdoǧan, Hakan

AU - Oflazer, Kemal

PY - 2005

Y1 - 2005

N2 - We face problems in large vocabulary continuous speech recognition for agglutinative languages, due to lack of coverage of all possible words. Since it is not enough to have a finite full-word lexicon, we may use sub-word units for recognition. Using sub-word lexicon units and developing language models based on these units solves the coverage problem. However, this results in increased acoustic confusability and shorter effective language model history length. We introduce new ways to choose lexicon units and we incorporate linguistic constraints into a statistical language model developed with the new units. We represent both the statistical language model and linguistic constraints as weighted finite state machines (WFSM) and combine them to obtain a novel language model. We study the performance of the new language model and show that it achieves 3% relative reduction in word error rate when used in recognizing a test-set of 2151 words.

AB - We face problems in large vocabulary continuous speech recognition for agglutinative languages, due to lack of coverage of all possible words. Since it is not enough to have a finite full-word lexicon, we may use sub-word units for recognition. Using sub-word lexicon units and developing language models based on these units solves the coverage problem. However, this results in increased acoustic confusability and shorter effective language model history length. We introduce new ways to choose lexicon units and we incorporate linguistic constraints into a statistical language model developed with the new units. We represent both the statistical language model and linguistic constraints as weighted finite state machines (WFSM) and combine them to obtain a novel language model. We study the performance of the new language model and show that it achieves 3% relative reduction in word error rate when used in recognizing a test-set of 2151 words.

UR - http://www.scopus.com/inward/record.url?scp=33846645022&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33846645022&partnerID=8YFLogxK

U2 - 10.1109/SIU.2005.1567632

DO - 10.1109/SIU.2005.1567632

M3 - Conference contribution

AN - SCOPUS:33846645022

SN - 0780392396

SN - 9780780392397

VL - 2005

SP - 111

EP - 114

BT - Proceedings of the IEEE 13th Signal Processing and Communications Applications Conference, SIU 2005

ER -