GU team’s research wins the Best Student Paper Award

The paper on ‘Empirical Analysis of Oral and Nasal Vowels of Konkani’ won at the 10th Language and Technology Conference (LTC 2023) in Poland. NT KURIOCITY gets you the details

RAMANDEEP KAUR

Konkani is a highly nasalised language which makes it unique among Indo-Aryan languages. The research paper, ‘Empirical Analysis of Oral and Nasal Vowels of Konkani’ investigates the acoustic-phonetic properties of Konkani oral and nasal vowels.

The research by Swapnil Fadte (assistant professor, Computer Science and Technology Discipline, Goa Business School- Goa University), Edna Vaz (Government College of Arts Science and Commerce Quepem, and Department of Linguistics, University of Mumbai), Atul Kumar Ojha (adjunct lecturer and postdoctoral fellow at the College of Science and Engineering, University of Galway, Ireland), Ramdas Karmali (assistant professor at Goa University) and Jyoti D. Pawar (professor and dean of Goa Business School) was recently presented by Fadte at the 10th Language and Technology Conference (LTC 2023): Human Language Technologies as a Challenge for Computer Science and Linguistics held in Poznan, Poland.

For this study, speech samples from six speakers (three male and three female) were collected. A total of 74 unique sentences were used as part of the recording script, 37 each for oral and nasal vowels, respectively. The final data set consisted of 1,135 vowel phonemes. A comparative plot of Konkani oral and nasal vowels was presented with an experimental result and formant analysis.

The paper which won them the Best Student Paper Award at the conference is helpful for the linguistic research on vowels and speech synthesis systems specific to the Konkani language. The research took around a year. The team did it in phases. First phase was to identify recording scripts, later they did recording of speakers and after that data cleaning and data verification was done by Vaz, who is a linguist. Once the data was verified, acoustic properties were extracted from speech data.

Later, the data was analysed to provide comparative study of oral and nasal vowels. Finally, the results were compiled and communicated at the conference.

Fadte, who is also a research scholar pursuing his Ph.D in the area of Automatic Speech Recognition (ASR), says, “This paper is a collaborative work which was done as an extension work on ‘Acoustic Analysis of Vowels in Konkani’ which I have previously worked on. With that study we actually read a few classifications which were given by linguists over time. However there was no agreement on the exact number of vowels as a part of Konkani language. So we classified the vowels in Konkani through acoustic study. With this paper we were able to prove that nasalisation is phonemic in Konkani.”

This means that the oral and nasal counterparts for the vowels when they occur in any word, changes the meaning of that word. For example, an oral sound ‘aa’ as in vaat (वाट, ‘way; path’. fem.sg.) means way or rosto. Whereas when you add nasalisation in ‘aa’ it becomes ‘aaM’ and the word becomes vaant (वांट, ‘grind’. v. 2p.sg.imperative’) which changes the meaning.

Thus, through this work they have identified close vowel pairs for all vowel sounds in Konkani. “We have done an acoustic study on it and what we have proved is that all oral sounds can be nasalised. We have shown that vowels occupy different positions in format space and it proves that nasalisation is phonemic which is quite unique for a language like Konkani and not common in most other Indo-Aryan languages,” he adds.

He mentions that they have not looked at all phones in language but have only worked on vowel phones. “The classification we took is from our previous work where we have classified all Konkani vowels. In this work we have identified the nasal counterpart of oral vowels and then we have identified word pairs like (शी (‘ugh’ excl.) and शीं (cold)) which are quite close in pronunciation.”

The phonemic property of nasalisation was known but never tested through experiment before. And with this research the team was able to prove the hypothesis that oral sounds can be nasalised and the nasalisation is phonemic.

Fadte tells us that most of the linguists agreed that Konkani vowels can be nasalised and nasalisation changes the meaning. “They gave some examples for few vowels but we never found examples for all vowels in any literature. Thus, we took this as a challenge and identified closed word pairs where just a single nasalisation property applied on the phoneme which is the vowel phoneme in that word which changes the meaning. This was an extensive study and we were fortunate to have a linguist as one of the co-authors in this work.”

Currently, Fadte in collaboration with the other authors is working on a phonetic dictionary for Konkani language, in order to provide accurate pronunciation of Konkani words.