Building Speech Recognition Systems for Language Documentation: The CoEDL Endangered Language Pipeline and Inference System (ELPIS)

Foley, Ben; Arnold, Josh; Coto-Solano, Rolando; Durantin, Gautier; Ellison, T. Mark; Esch, Daan van; Heath, Scott; Kratochvíl, František; Maxwell-Smith, Zara; Nash, Douglas B.; Olsson, Ola; Richards, Mark; San, Nay; Stoakes, Hywel; Thieberger, Nick; Wiles, Janet

doi:10.21437/sltu.2018-43

Cited by 107 publications

(16 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Elpis 1 is a tool created to allow language workers with minimal computational experience to build their own speech recognition models and automatically transcribe audio (Foley et al, 2018. Elpis uses the Kaldi 2 automatic speech recognition (ASR) toolkit (Povey et al, 2011) as its backend.…”

Section: Introductionmentioning

confidence: 99%

User-Friendly Automatic Transcription of Low-Resource Languages: Plugging ESPnet into Elpis

et al. 2021

View full text Add to dashboard Cite

This paper reports on progress integrating the speech recognition toolkit ESPnet into Elpis, a web front-end originally designed to provide access to the Kaldi automatic speech recognition toolkit. The goal of this work is to make end-to-end speech recognition models available to language workers via a user-friendly graphical interface. Encouraging results are reported on (i) development of an ESPnet recipe for use in Elpis, with preliminary results on data sets previously used for training acoustic models with the Persephone toolkit along with a new data set that had not previously been used in speech recognition, and (ii) incorporating ESPnet into Elpis along with UI enhancements and a CUDA-supported Dockerfile.

show abstract

Section: Introductionmentioning

confidence: 99%

User-Friendly Automatic Transcription of Low-Resource Languages: Plugging ESPnet into Elpis

et al. 2021

View full text Add to dashboard Cite

show abstract

“…It was reported that the performance of MFCC based systems decreases with decreasing frame size [9,10]. Classifiers like Hidden Markov Model (HMM) [12], vector quantization (VQ) [6], support vector machine (SVM) [3,13,14], artificial neural network (ANN) [15,16], and Gaussian mixture model (GMM) [15][16][17] have been reported to model feature vectors in SLID systems. One of the simplest techniques used for the SLID system is GMM-UBM.…”

Section: Introductionmentioning

confidence: 99%

“…In this method, maximum likelihood estimation is used to train the language model, and maximum a posterior (MAP) estimation is used to adapt the UBM model. The speech sample is a series of the independent spectral feature vector, and GMM mathematically models these features with UBM adaption known as GMM-UBM supervectors carries spectral characteristics [10][11][12][13][14][15][16][17][18]. These features are adapted to UBM using the MAP estimation algorithm to obtain utterance-based GMM [19].…”

Section: Introductionmentioning

confidence: 99%

A GMM supervector approach for spoken Indian language identification for mismatch utterance length

Bakshi

Kopparapu²

2021

Bulletin EEI

View full text Add to dashboard Cite

Gaussian mixture model-universal background model (GMM UBM) supervectors are used to identify spoken Indian languages. The supervectors are calculated from short-time MFCC, its first and sec derivatives. The UBM builds a generalized Indian language model, and mean adaptation transforms it to a duration normalized language-specific GMM. Multi-class support vector machine and artificial neural network classifiers are used to identify language labels from the supervectors. Experimental evaluations are performed using 30 sec speech utterances from nine Indian languages comprised five Indo-Aryan and four Dravidian languages, extracted from all India radio broadcast news data-set. Eight smaller duration data-sets were manually derived to study the effect of training and test duration mismatch. In mismatch conditions, identification accuracy decreases with a decrease in test and train utterance duration. Investigations showed that the 32-mixture model with ANN classifier has optimal performance.

show abstract

“…Indigenous and other minority languages usually have few transcribed audio recordings, and so adapting data-hungry ASR algorithms to assist in their documentation is an active area of research (Besacier et al, 2014;Jimerson and Prud'hommeaux, 2018;Michaud et al, 2019;Foley et al, 2018;Gupta and Boulianne, 2020b,a;Zahrer et al, 2020;Thai et al, 2019;Li et al, 2020;Zevallos et al, 2019;Matsuura et al, 2020;Levow et al, 2021). This paper will examine an element that might appear obvious at first, but one where the literature is "inconclusive" (Adams, 2018), and which can have major consequences in performance: How should tones be transcribed when dealing with extremely low-resource languages?…”

Section: Introductionmentioning

confidence: 99%

Explicit Tone Transcription Improves ASR Performance in Extremely Low-Resource Languages: A Case Study in Bribri

Coto-Solano¹

2021

Proceedings of the First Workshop on Natural Language Processing for Indigenous Languages of the Americas

Self Cite

View full text Add to dashboard Cite

Linguistic tone is transcribed for input into ASR systems in numerous ways. This paper shows a systematic test of several transcription styles, using as an example the Chibchan language Bribri, an extremely low-resource language from Costa Rica. The most successful models separate the tone from the vowel, so that the ASR algorithms learn tone patterns independently. These models showed improvements ranging from 4% to 25% in character error rate (CER), and between 3% and 23% in word error rate (WER). This is true for both traditional GMM/HMM and end-to-end CTC algorithms. This paper also presents the first attempt to train ASR models for Bribri. The best performing models had a CER of 33% and a WER of 50%. Despite the disadvantage of using hand-engineered representations, these models were trained on only 68 minutes of data, and therefore show the potential of ASR to generate further training materials and aid in the documentation and revitalization of the language. ResumenTranscribir el tono de forma explícita mejora el rendimiento del reconocimiento de voz en idiomas extremadamente bajos en recursos: Un estudio de caso en bribri. Hay numerosas maneras de transcribir el tono lingüístico a la hora de proveer los datos de entrenamiento a los sistemas de reconocimiento de voz. Este artículo presenta un experimento sistemático de varias formas de transcripción usando como ejemplo la lengua chibcha bribri, una lengua de Costa Rica extremadamente baja en recursos. Los modelos más exitosos fueron aquellos en que el tono aparece separado de la vocal de tal forma que los algoritmos pudieran aprender los patrones tonales por separado. Estos modelos mostraron mejoras de entre 4% y 26% en el error de caracteres (CER), y de entre 3% y 25% en el error de palabras (WER). Esto se observó tanto en los algoritmos GMM/HMM como en los algoritmos CTC de secuenciaa-secuencia. Este artículo también presenta el primer intento de entrenar modelos de reconocimiento de voz en bribri. Los mejores modelos tuvieron un CER de 33% y un WER de 50%. A pesar de la desventaja de usar representaciones diseñadas a mano, estos modelos se entrenaron con solo 68 minutos de datos y muestran el potencial para generar más materiales de entrenamiento, así como de ayudar con la documentación y revitalización de la lengua.

show abstract

Building Speech Recognition Systems for Language Documentation: The CoEDL Endangered Language Pipeline and Inference System (ELPIS)

Cited by 107 publications

References 1 publication

User-Friendly Automatic Transcription of Low-Resource Languages: Plugging ESPnet into Elpis

User-Friendly Automatic Transcription of Low-Resource Languages: Plugging ESPnet into Elpis

A GMM supervector approach for spoken Indian language identification for mismatch utterance length

Explicit Tone Transcription Improves ASR Performance in Extremely Low-Resource Languages: A Case Study in Bribri

Contact Info

Product

Resources

About