Toward an End-to-End Voice to Sign Recognition for Dialect Moroccan Language

Allak, Anass; Benelallam, Imade; Habbouza, Hamdi; Amallah, Mohamed

doi:10.1007/978-3-030-94188-8_24

Cited by 7 publications

(6 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors also explored works related to Moroccan Darija resources, such as lexicons, corpora, and dialect identification, emphasizing the importance of annotated corpora for further research in this area.  In 2021, the Dvoice dataset 1.0, developed by SI2M labs in collaboration with AIOX Lab [11] [12] proposed a solution based on the state-of-the-art architecture named deep speech 2 by Baidu. They conducted tests on 24 hours of speech data (DarSpeech) and achieved promising results, with a word error rate (WER) of 22.7% and a character error rate of 6.03%.…”

Section: Related Workmentioning

confidence: 99%

MDVC corpus: empowering Moroccan Darija speech recognition

Ahmed,

Abdellah

2024

IJEECS

View full text Add to dashboard Cite

Automatic speech recognition (ASR) technology has significantly transformed human-machine interactions, but it remains limited in its representation of diverse languages and dialects. Moroccan Darija, the lively Moroccan dialect, has long been underrepresented in the realm of language technology. To address this gap, we present a novel corpus of audio files accompanied by meticulously transcribed Moroccan Darija speech. The corpus comprises 1,000 hours of diverse content, featuring multiple Moroccan accents, extracted from 80 YouTube channels. To standardize the representation of Moroccan Darija in our corpus, we made efforts to establish consistent writing norms and conventions. In addition to the dataset creation, we applied fine-tuning using the Wav2Vec2 model on the Moroccan Darija voice corpus (MDVC) dataset achieving a remarkable word error rate (WER) of 9%. This article discusses the current state of Moroccan Darija research, highlighting the scarcity of resources and the need for robust ASR systems. Our contribution offers a valuable resource for researchers and developers, and by standardizing the Darija language, we strive to improve ASR system for this low resource language.

show abstract

Section: Related Workmentioning

confidence: 99%

MDVC corpus: empowering Moroccan Darija speech recognition

Ahmed,

Abdellah

2024

IJEECS

View full text Add to dashboard Cite

show abstract

“…The dataset was highly unbalanced, and the original data for languages such as Darija, Yoruba, and Sudanese were extremely low. Thereafter, we conduct experiments based on augmented data using the augmentation method proposed in [27], which adds a perturbation, augmentation lobe and also corrupts speech features by adding a frequency and a time dropout.…”

Section: Language Choices and Datasetsmentioning

confidence: 99%

“…We use the ALFFA Public [24][25][26] dataset for the Amharic, Fongbe, Swahili and Wolof languages (see Table 2). For Darija, the DVoice dataset was used [27], a dataset created using pseudo-labelling and data augmentation techniques. The Sudanese dataset was collected from Youtube videos then labeled manually.…”

Section: Language Choices and Datasetsmentioning

confidence: 99%

“…Many solutions can be tested, such as adding new data for languages with low training datasets or performing data augmentation that corrupts the audio by adding perturbations, frequency, time dropouts, and augmentation lobes. Here is where the data augmentation method proposed in [27] which we applied to the Sudanese, Darija, and Yoruba classes. According to our previous experiments, the MFCCs approach performs well; therefore, we focused only on the augmented data.…”

Section: Language Identificationmentioning

confidence: 99%

See 1 more Smart Citation

Multilingual Speech Recognition Initiative for African Languages

Mohamed¹,

Benelallam²,

Allak³

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

This paper summarizes a speech recognition initiative for African languages. More precisely, we propose innovative approaches that address the low-resource property of these languages. For both monolingual and multilingual systems, our methods rely on self-supervised pre-trained models for multiple languages. We tested our method on seven African languages and dialects: Amharic, Darija, Fongbe, Sudanese, Swahili, Wolof, and Yoruba. We first trained monolingual models that were used as baselines, and then proposed proof-of-concepts for systems that handle multiple languages. Our multilingual systems were based on three scenarios: (a) we trained a single model by concate-nating the multilingual corpora; (b) we discussed this first model by testing another joint model that predicts the spoken language using language-specific tokens before the text transcription; and (c) we fed a one-hot encoder vector to the latent feature extractions before training the single model and for inference. For this purpose, a language identification model is required. We also investigated the impact of lexical ambiguity by removing diacritics from text in some languages.

show abstract

“…In the paper [1] Authors presents building an automated voice to sign system based on pose estimation but generating more own animations for not available poses can improvise the system. In paper [2] authors discussed on text to sign conversion for Arabic sign language and it can be extended to other languages with more vocabulary.…”

Section: Related Workmentioning

confidence: 99%

Effective Communication Between Differently Abled And Normal People Using Speech To Sign Translation System

N¹

2022

Journal of Pharmaceutical Negative Results

View full text Add to dashboard Cite

There are very less way of communicating between a differently abled person i.e. deaf and dumb also called as hearing impaired and normal people. Communications from deaf and dumb to normal people are done in many ways but from normal people to hearing impaired people have not been much. What we say cannot be understood by the deaf and dumb people so a system should be made so that a deaf and dumb person can understand the voice of normal people. The voice will be converted to sign language so that deaf and dumb people can easily understand. Machine learning techniques will be used to implement this with the help of Google API for voice recognition and then the Natural Language Processing is used. The voice will be broken into sentences or words and then the context of natural language processing will come into picture. The text will be compared with the gifs generated, if found the gifs will be shown as output or else letter by letter mapping will be done so the hearing impaired people can understand the voice through sign language. This is very important for the hearing impaired people. This helps hearing impaired people a lot because they can easily understand what the normal people are telling and for their daily life it will be very helpful. The system is tested with the number of data samples. As data samples increases, the accuracy of the model is said to have gone up as well.

show abstract

Toward an End-to-End Voice to Sign Recognition for Dialect Moroccan Language

Cited by 7 publications

References 14 publications

MDVC corpus: empowering Moroccan Darija speech recognition

MDVC corpus: empowering Moroccan Darija speech recognition

Multilingual Speech Recognition Initiative for African Languages

Effective Communication Between Differently Abled And Normal People Using Speech To Sign Translation System

Contact Info

Product

Resources

About