Wolwala : deep pashto text-to-speech | |
| Author | Safi, Abdul Rahman |
| Call Number | AIT RSPR no.IM-19-06 |
| Subject(s) | Machine learning Natural language processing (Computer science) Speech synthesis |
| Note | A research submitted in partial fulfillment of the requirements for the degree of Master of Science in Information Management, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Research studies project report ; no. IM-19-06 |
| Abstract | Nowadays, communication plays a vital role in our highly connected world. It is practically impossible to live without information exchange. One form of information exchange is through text to speech, which benefits people in various domains and numerous applications. Text to speech combined with speech-to-text and machine translation is proven effective for cross-cultural communication, enabling an individual to communicate through his native language with speakers of other languages. Wolwala deep Pashto text-to-Speech is the text-to-speech system developed specifically for the Pashto language. Pashto is the native language of nearly 60 million people, mainly living across Afghanistan and Pakistan. Themainchallengesincludedtheunavailibilityofthedatasetandmanylanguage-dependent componentssuchastokenizers,partofspeechtaggers,phrasebreakers,andphonemicdictionary. Wolwalaaddressedthesechallengeswithanunsupervisedlearningandsomelanguageindependent techniques. The developed text-to-speech system is based on RNN-LSTM architecture. Finaly, the system is evaluated employing both subjective and objective measures. The results were excellent. However, better results could be achieved with more data and Pashto-specific components. |
| Year | 2019 |
| Corresponding Series Added Entry | Asian Institute of Technology. Research studies project report ; no. IM-19-06 |
| Type | Research Study Project Report (RSPR) |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information Management (IM) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Mongkol Ekpanyapong;Phan Minh Dung; |
| Scholarship Donor(s) | Ministry of Higher Education (MoHE), Afghanistan; |
| Degree | Research studies project report (M. Sc.) - Asian Institute of Technology, 2019 |