Research Projects

Development of technology solutions with Indian languages is making significant progress in the past decade. However, the North East Indian languages have not found a space in the developments primarily due to two reasons: a) several of the North Eastern Tibeto-Burman languages are low resourced and b) most of the languages are not scheduled languages. Apart from the lack of language technology development, the development of test batteries and speech therapeutic tools is also slow in these languages. In terms of speech pathologies, hypernasal speech emerging due to cleft lip and palate deformation is very common in the North East Indian languages. While the cause is not determined, it is estimated that a single clinic run in Guwahati performs nearly 5000 surgeries in a year to reconstruct cleft lip and palate. In order to estimate the severity of hypernasal speech, the normative nasalance score of a population needs to be determined. Secondly, as nasal properties are language specific, acoustic analysis of nasals in languages is essential for diagnostic and language technology solutions such as, Text to Speech (TTS) and automatic Speech recognition (ASR). In order to aid language technology development and speech pathology, in depth analysis of linguistic populations and creation of knowledge base for the acoustic properties of the languages is required. Moreover, special attention needs to be paid to the unique speech sounds in the linguistic populations. Hence, in the current project, we plan to study the unique nasal sounds of three Tibeto-Burman languages, spoken in North East India, namely, Angami, Khezha and Mizo. While most of the languages in the world have voiced nasals, these three languages also have a voiceless inventory of nasals. This is a highly rare property in spoken languages, found only in 3.9% of the world’s languages. However, the voiceless nasals are in abundance in the Tibeto-Burman languages of North East India. Hence, in order to create effective language technology and speech therapeutic tools, the fine grained acoustic properties of the unique nasals in these languages need to be explored. The outcomes of this project will further the scientific knowledge about the acoustic characteristics of the nasals in general and voiceless nasals in particular. Secondly, the creation of the nasal test battery and the determination of the normative nasalance will provide groundtruth for speech therapists to determine hypernasality and hyponasality in the three languages. This may be extended to other closely related languages that have voicing contrasts in nasals. Thirdly, machine learning based classification of nasality explored in this work will facilitate language technology-based solutions, such as ASR and TTS. Finally, the project is a step towards scientifically documenting the three under-resourced languages, namely, Angami, Khezha and Mizo, that have some rare phonological properties.