Using different acoustic, lexical and language modeling units for ASR of an under-resourced language - Amharic

By Martha Yifiru Tachbelie, S. Abate, Laurent Besacier. 2014

49 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications, Computational Linguistics & Semantics, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. James R. Glass
2. Wei-Ning Hsu
3. Mohit Bansal
4. Hung-yi Lee
5. Graham Neubig
6. Michael Auli
7. Yoshua Bengio
8. Abdel-rahman Mohamed

In Computer Science Applications

1. Laurent Besacier
2. Wei-Ning Hsu

In Computational Linguistics & Semantics

1. Chris Dyer
2. Phil Blunsom
3. Kazuya Kawakami
4. Aäron van den Oord
5. Luyu Wang
6. Laurent Besacier
7. Tanja Schultz
8. Karen Livescu

In Speech & Audio

1. Wei-Ning Hsu
2. Abdel-rahman Mohamed
3. Alexei Baevski
4. Hung-yi Lee
5. Michael Auli
6. James R. Glass
7. Alexis Conneau
8. Jinyu Li

In Signal Processing

1. Wei-Ning Hsu
2. Abdel-rahman Mohamed
3. Hung-yi Lee
4. James R. Glass
5. Alexei Baevski
6. Jinyu Li
7. Dong Yu
8. Shinji Watanabe

Citations in the corpus

• Unsupervised Speech Recognition (2021, methodology): “We experiment with the Swahili data from the ALFFA project (Gelas et al., 2012; Abate et al., 2005; Tachbelie et al., 2014) which is read speech.”
• Learning Robust and Multilingual Speech Representations (2020, positive): “We also train ASR models on 25 languages, including low-resource languages (e.g. Amharic, Fongbe, Swahili, Wolof), and show that our representations significantly outperform both standard features and those pretrained only on clean English data in the language transfer setup.”
• Self-Supervised Speech Representation Learning: A Review (2022, background): “The ALFFA project3 collects speech of African languages to promote the development of speech technologies in Africa, and [123] leverages four African languages collected in the project for evaluation: Amharic [241], Fongbe [242], Swahili [243], and”
• Collecting Resources in Sub-Saharan African Languages for Automatic Speech Recognition: a Case Study of Wolof (2016, methodology): “The Amharic system was retrained from the corpus described in (Tachbelie et al., 2014) and represents about 20 hours of speech data, while the testing set represents about 2 hours.”
• Speech technologies for african languages: example of a multilingual calculator for education (2015, methodology): “The Amharic system was retrained from the corpus described in [4].”
• Automatic speech recognition for under-resourced languages: A survey (2014, methodology): “A less radical departure from the standard model uses Hidden Markov Models to model syllables rather than phonemes [Tachbelie, 2012 ; Tachbelie, 2013] ; – in this case, the reduction in model parameters results from the fact that context dependencies are generally less important for syllable models.”