Towards phoneme inventory discovery for documentation of unwritten languages

Müller, Markus; H., Franke, Jörg K.; Waibel, Alex; Stüker, Sebastian

doi:10.1109/icassp.2017.7953148

Cited by 14 publications

(14 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Articulatory phonetics describes the mechanism of speech production such as the manner of articulation and placement of articulation, and it tends to describe phones using discrete features such as voiced, bilabial (made with the two lips) and fricative. These articulatory features have been shown to be useful in speech recognition (Kirchhoff 1998;Stüker et al 2003b;Müller et al 2017), and are a good choice for attributes for our purpose. We provide some categories of articulatory attributes below.…”

Section: Articulatory Attributesmentioning

confidence: 99%

“…Articulatory features have been shown to be useful in speech recognition under several situation. Specifically, articulatory features has been used to improve robustness under noisy and reverberant environment (Kirchhoff 1998), compensate for crosslingual variability (Stüker et al 2003b), improve word error rate in multilingual models (Stüker et al 2003a), be beneficial for low resource languages (Müller, Stüker, and Waibel 2016), detecting spoken words (Prabhavalkar et al 2013), clustering phoneme-like units for unwritten languages (Müller et al 2017), recognizing unseen languages (Siniscalchi et al 2011), developing phonological vocoder (Cernak and Garner 2016). There are also some attempts to predict articulatory features or distributions for clinical usages (Jiao, Berisha, and Liss 2017;Vásquez-Correa et al 2019), but they do not provide a model to predict unseen phonemes.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Towards Zero-Shot Learning for Automatic Phonemic Transcription

Dalmia

Mortensen

et al. 2020

AAAI

View full text Add to dashboard Cite

Automatic phonemic transcription tools are useful for low-resource language documentation. However, due to the lack of training sets, only a tiny fraction of languages have phonemic transcription tools. Fortunately, multilingual acoustic modeling provides a solution given limited audio training data. A more challenging problem is to build phonemic transcribers for languages with zero training data. The difficulty of this task is that phoneme inventories often differ between the training languages and the target language, making it infeasible to recognize unseen phonemes. In this work, we address this problem by adopting the idea of zero-shot learning. Our model is able to recognize unseen phonemes in the target language without any training data. In our model, we decompose phonemes into corresponding articulatory attributes such as vowel and consonant. Instead of predicting phonemes directly, we first predict distributions over articulatory attributes, and then compute phoneme distributions with a customized acoustic model. We evaluate our model by training it using 13 languages and testing it using 7 unseen languages. We find that it achieves 7.7% better phoneme error rate on average over a standard multilingual model.

show abstract

Section: Articulatory Attributesmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Towards Zero-Shot Learning for Automatic Phonemic Transcription

Dalmia

Mortensen

et al. 2020

AAAI

View full text Add to dashboard Cite

show abstract

“…When we are unable to hear what linguists and speakers would tell us, we must fall back on unsupervised methods, like discovering a phone inventory from speech alone (Kempton and Moore, 2014;Vetter et al, 2016;Adda et al, 2016;Müller et al, 2017), or segmenting phone sequences into "words" (Johnson and Goldwater, 2009;Elsner et al, 2013). Yet after centuries of colonisation, missionary endeavours, and linguistic fieldwork, all languages have been identified and classified.…”

Section: Technology Saving Languagesmentioning

confidence: 99%

Decolonising Speech and Language Technology

Bird¹

2020

Proceedings of the 28th International Conference on Computational Linguistics

View full text Add to dashboard Cite

After generations of exploitation, Indigenous people often respond negatively to the idea that their languages are data ready for the taking. By treating Indigenous knowledge as a commodity, speech and language technologists risk disenfranchising local knowledge authorities, reenacting the causes of language endangerment. Scholars in related fields have responded to calls for decolonisation, and we in the speech and language technology community need to follow suit, and explore what this means for our practices that involve Indigenous languages and the communities who own them. This paper reviews colonising discourses in speech and language technology, and suggests new ways of working with Indigenous communities, and seeks to open a discussion of a postcolonial approach to computational methods for supporting language vitality.

show abstract

“…This is taken a step further in [5], which derives a phoneme-like segmental representation of the Articulatory Features. This approach is again advanced in [6], where BiLSTMs are used to first identify segment boundaries before extracting and clustering the Articulatory Features.…”

Section: Related Workmentioning

confidence: 99%

Unsupervised Phonetic and Word Level Discovery for Speech to Speech Translation for Unwritten Languages

Hillis¹,

Kumar²,

Black

2019

Interspeech 2019

View full text Add to dashboard Cite

We experiment with unsupervised methods for deriving and clustering symbolic representations of speech, working towards speech-to-speech translation for languages without regular (or any) written representations. We consider five low-resource African languages, and we produce three different segmental representations of text data for comparisons against four different segmental representations derived solely from acoustic data for each language. The text and speech data for each language comes from the CMU Wilderness dataset introduced in [1], where speakers read a version of the New Testament in their language. Our goal is to evaluate the translation performance not only of acoustically derived units but also of discovered sequences or "words" made from these units, with the intuition that such representations will encode more meaning than phones alone. We train statistical machine translation models for each representation and evaluate their outputs on the basis of BLEU-1 scores to determine their efficacy. Our experiments produce encouraging results: as we cluster our atomic phonetic representations into more word-like units, the amount information retained generally approaches that of the actual words themselves.

show abstract

Towards phoneme inventory discovery for documentation of unwritten languages

Cited by 14 publications

References 17 publications

Towards Zero-Shot Learning for Automatic Phonemic Transcription

Towards Zero-Shot Learning for Automatic Phonemic Transcription

Decolonising Speech and Language Technology

Unsupervised Phonetic and Word Level Discovery for Speech to Speech Translation for Unwritten Languages

Contact Info

Product

Resources

About