Multilingual Speech Recognition for Turkic Languages
The primary aim of this study was to contribute to the development of multilingual automatic speech recognition for lower-resourced Turkic languages. Ten languages—Azerbaijani, Bashkir, Chuvash, Kazakh, Kyrgyz, Sakha, Tatar, Turkish, Uyghur, and Uzbek—were considered. A total of 22 models were devel...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-01-01
|
| Series: | Information |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2078-2489/14/2/74 |
| _version_ | 1797620346123190272 |
|---|---|
| author | Saida Mussakhojayeva Kaisar Dauletbek Rustem Yeshpanov Huseyin Atakan Varol |
| author_facet | Saida Mussakhojayeva Kaisar Dauletbek Rustem Yeshpanov Huseyin Atakan Varol |
| author_sort | Saida Mussakhojayeva |
| collection | DOAJ |
| description | The primary aim of this study was to contribute to the development of multilingual automatic speech recognition for lower-resourced Turkic languages. Ten languages—Azerbaijani, Bashkir, Chuvash, Kazakh, Kyrgyz, Sakha, Tatar, Turkish, Uyghur, and Uzbek—were considered. A total of 22 models were developed (13 monolingual and 9 multilingual). The multilingual models that were trained using joint speech data performed more robustly than the baseline monolingual models, with the best model achieving an average character and word error rate reduction of 56.7%/54.3%, respectively. The results of the experiment showed that character and word error rate reduction was more likely when multilingual models were trained with data from related Turkic languages than when they were developed using data from unrelated, non-Turkic languages, such as English and Russian. The study also presented an open-source Turkish speech corpus. The corpus contains 218.2 h of transcribed speech with 186,171 utterances and is the largest publicly available Turkish dataset of its kind. The datasets and codes used to train the models are available for download from our GitHub page. |
| first_indexed | 2024-03-11T08:39:57Z |
| format | Article |
| id | doaj.art-9bfe34fe2bca4bedb0dc44d009286b3d |
| institution | Directory Open Access Journal |
| issn | 2078-2489 |
| language | English |
| last_indexed | 2024-03-11T08:39:57Z |
| publishDate | 2023-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Information |
| spelling | doaj.art-9bfe34fe2bca4bedb0dc44d009286b3d2023-11-16T21:11:58ZengMDPI AGInformation2078-24892023-01-011427410.3390/info14020074Multilingual Speech Recognition for Turkic LanguagesSaida Mussakhojayeva0Kaisar Dauletbek1Rustem Yeshpanov2Huseyin Atakan Varol3Institute of Smart Systems and Artificial Intelligence (ISSAI), Nazarbayev University, Astana 010000, KazakhstanInstitute of Smart Systems and Artificial Intelligence (ISSAI), Nazarbayev University, Astana 010000, KazakhstanInstitute of Smart Systems and Artificial Intelligence (ISSAI), Nazarbayev University, Astana 010000, KazakhstanInstitute of Smart Systems and Artificial Intelligence (ISSAI), Nazarbayev University, Astana 010000, KazakhstanThe primary aim of this study was to contribute to the development of multilingual automatic speech recognition for lower-resourced Turkic languages. Ten languages—Azerbaijani, Bashkir, Chuvash, Kazakh, Kyrgyz, Sakha, Tatar, Turkish, Uyghur, and Uzbek—were considered. A total of 22 models were developed (13 monolingual and 9 multilingual). The multilingual models that were trained using joint speech data performed more robustly than the baseline monolingual models, with the best model achieving an average character and word error rate reduction of 56.7%/54.3%, respectively. The results of the experiment showed that character and word error rate reduction was more likely when multilingual models were trained with data from related Turkic languages than when they were developed using data from unrelated, non-Turkic languages, such as English and Russian. The study also presented an open-source Turkish speech corpus. The corpus contains 218.2 h of transcribed speech with 186,171 utterances and is the largest publicly available Turkish dataset of its kind. The datasets and codes used to train the models are available for download from our GitHub page.https://www.mdpi.com/2078-2489/14/2/74automatic speech recognitionmultilingual speech recognitionTurkic languagestransfer learningCommon Voicebig data |
| spellingShingle | Saida Mussakhojayeva Kaisar Dauletbek Rustem Yeshpanov Huseyin Atakan Varol Multilingual Speech Recognition for Turkic Languages Information automatic speech recognition multilingual speech recognition Turkic languages transfer learning Common Voice big data |
| title | Multilingual Speech Recognition for Turkic Languages |
| title_full | Multilingual Speech Recognition for Turkic Languages |
| title_fullStr | Multilingual Speech Recognition for Turkic Languages |
| title_full_unstemmed | Multilingual Speech Recognition for Turkic Languages |
| title_short | Multilingual Speech Recognition for Turkic Languages |
| title_sort | multilingual speech recognition for turkic languages |
| topic | automatic speech recognition multilingual speech recognition Turkic languages transfer learning Common Voice big data |
| url | https://www.mdpi.com/2078-2489/14/2/74 |
| work_keys_str_mv | AT saidamussakhojayeva multilingualspeechrecognitionforturkiclanguages AT kaisardauletbek multilingualspeechrecognitionforturkiclanguages AT rustemyeshpanov multilingualspeechrecognitionforturkiclanguages AT huseyinatakanvarol multilingualspeechrecognitionforturkiclanguages |