Meta Learning Approach to Phone Duration Modeling
One of the essential prerequisites for achieving the naturalness of synthesized speech is the possibility of the automatic prediction of phone duration, due to the high importance of segmental duration in speech perception. In this paper we present a new phone duration prediction model for the Serbi...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Faculty of Mechanical Engineering in Slavonski Brod, Faculty of Electrical Engineering in Osijek, Faculty of Civil Engineering in Osijek
2018-01-01
|
| Series: | Tehnički Vjesnik |
| Subjects: | |
| Online Access: | https://hrcak.srce.hr/file/298283 |
| _version_ | 1797207643763245056 |
|---|---|
| author | Sandra Sovilj-Nikić Ivan Sovilj-Nikić Maja Marković |
| author_facet | Sandra Sovilj-Nikić Ivan Sovilj-Nikić Maja Marković |
| author_sort | Sandra Sovilj-Nikić |
| collection | DOAJ |
| description | One of the essential prerequisites for achieving the naturalness of synthesized speech is the possibility of the automatic prediction of phone duration, due to the high importance of segmental duration in speech perception. In this paper we present a new phone duration prediction model for the Serbian language using meta learning approach. Based on the data obtained from the analysis of a large speech database, we used a feature set of 21 parameters describing phones and their contexts. These include attributes related to the segmental identity, manner of articulation (for consonants), attributes related to phonological context, such as segment types and voicing values of neighboring phones, presence or absence of lexical stress, morphological attributes, such as part-of-speech, and prosodic attributes, such as phonological word length, the position of the segment in the syllable, the position of the syllable in a word, the position of a word in a phrase, phrase break level, etc. Phone duration model obtained using meta learning algorithm outperformed the best individual model by approximately 2,0% and 1,7% in terms of the relative reduction of the root-mean-squared error and the mean absolute error, respectively. |
| first_indexed | 2024-04-24T09:26:11Z |
| format | Article |
| id | doaj.art-daedaa8cd3404787a4bf7d0a3ab67df7 |
| institution | Directory Open Access Journal |
| issn | 1330-3651 1848-6339 |
| language | English |
| last_indexed | 2024-04-24T09:26:11Z |
| publishDate | 2018-01-01 |
| publisher | Faculty of Mechanical Engineering in Slavonski Brod, Faculty of Electrical Engineering in Osijek, Faculty of Civil Engineering in Osijek |
| record_format | Article |
| series | Tehnički Vjesnik |
| spelling | doaj.art-daedaa8cd3404787a4bf7d0a3ab67df72024-04-15T14:53:49ZengFaculty of Mechanical Engineering in Slavonski Brod, Faculty of Electrical Engineering in Osijek, Faculty of Civil Engineering in OsijekTehnički Vjesnik1330-36511848-63392018-01-0125385586010.17559/TV-20171002122930Meta Learning Approach to Phone Duration ModelingSandra Sovilj-Nikić0Ivan Sovilj-Nikić1Maja Marković2Iritel a.d. Beograd, Batajnički put 23, 11080 Beorgad, SerbiaUniversity of Novi Sad, Faculty of Technical Sciences, Trg Dositeja Obradovića 6, 21000 Novi Sad, SerbiaUniversity of Novi Sad, Faculty of Philosophy, Dr Zorana Đinđića 2, 21000 Novi Sad, SerbiaOne of the essential prerequisites for achieving the naturalness of synthesized speech is the possibility of the automatic prediction of phone duration, due to the high importance of segmental duration in speech perception. In this paper we present a new phone duration prediction model for the Serbian language using meta learning approach. Based on the data obtained from the analysis of a large speech database, we used a feature set of 21 parameters describing phones and their contexts. These include attributes related to the segmental identity, manner of articulation (for consonants), attributes related to phonological context, such as segment types and voicing values of neighboring phones, presence or absence of lexical stress, morphological attributes, such as part-of-speech, and prosodic attributes, such as phonological word length, the position of the segment in the syllable, the position of the syllable in a word, the position of a word in a phrase, phrase break level, etc. Phone duration model obtained using meta learning algorithm outperformed the best individual model by approximately 2,0% and 1,7% in terms of the relative reduction of the root-mean-squared error and the mean absolute error, respectively.https://hrcak.srce.hr/file/298283machine learningmeta learning algorithmphone duration modelsynthesized speech |
| spellingShingle | Sandra Sovilj-Nikić Ivan Sovilj-Nikić Maja Marković Meta Learning Approach to Phone Duration Modeling Tehnički Vjesnik machine learning meta learning algorithm phone duration model synthesized speech |
| title | Meta Learning Approach to Phone Duration Modeling |
| title_full | Meta Learning Approach to Phone Duration Modeling |
| title_fullStr | Meta Learning Approach to Phone Duration Modeling |
| title_full_unstemmed | Meta Learning Approach to Phone Duration Modeling |
| title_short | Meta Learning Approach to Phone Duration Modeling |
| title_sort | meta learning approach to phone duration modeling |
| topic | machine learning meta learning algorithm phone duration model synthesized speech |
| url | https://hrcak.srce.hr/file/298283 |
| work_keys_str_mv | AT sandrasoviljnikic metalearningapproachtophonedurationmodeling AT ivansoviljnikic metalearningapproachtophonedurationmodeling AT majamarkovic metalearningapproachtophonedurationmodeling |