Large Vocabulary Read Speech Corpora for Four Ethiopian Languages: Amharic, Tigrigna, Oromo and Wolaytta

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4167–4171 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4167 Large Vocabulary Read Speech Corpora for Four Ethiopian Languages: Amharic, Tigrigna, Oromo and Wolaytta Solomon Teferra Abate, Martha Yifiru Tachbelie, Michael Melese, Hafte Abera, Tewodros Abebe, Wondwossen Mulugeta, Yaregal Assabie, Million Meshesha, Solomon Atinafu, Binyam Ephrem Addis Ababa University solomon.teferra, martha.yifiru, michael.melese, hafte.abera, tewodros.abebe, wondwossen.mulugeta, yaregal.assabie, million.meshesha, solomon.atnafu, binyam.ephrem@aau.edu.et Abstract Automatic Speech Recognition (ASR) is one of the most important technologies to support spoken communication in modern life. However, its development benefits from large speech corpus. The development of such a corpus is expensive and most of the human languages, including the Ethiopian languages, do not have such resources. To address this problem, we have developed four large (about 22 hours) speech corpora for four Ethiopian languages: Amharic, Tigrigna, Oromo and Wolaytta. To assess usability of the corpora for (the purpose of) speech processing, we have developed ASR systems for each language. In this paper, we present the corpora and the baseline ASR systems we have developed. We have achieved word error rates (WERs) of 37.65%, 31.03%, 38.02%, 33.89% for Amharic, Tigrigna, Oromo and Wolaytta, respectively. This results show that the corpora are suitable for further investigation towards the development of ASR systems. Thus, the research community can use the corpora to further improve speech processing systems. From our results, it is clear that the collection of text corpora to train strong language models for all of the languages is still required, especially for Oromo and Wolaytta. Keywords: Speech Corpus, Amharic, Tigrigna, Oromo, Wolaytta 1. Introduction Natural Language Processing (NLP) applications for many human languages is becoming a required factor for the ful- fillment of equity in the information access because one should not be discriminated from the use of the commu- nication technologies just because of not speaking the tech- nologically favored languages. NLP applications are use- ful in facilitating human-machine and even human-human communications that are inter-mediated by machines. One of the NLP applications which facilitates human-machine and human-human communication is Automatic Speech Recognition (ASR). ASR is the conversion of speech to the corresponding tex- tual representation. It enables a digital machine to extract the information conveyed in speech as a sequence of words. The resulting sequence of words can be sent to any search engine or can be used as the transcription of any human speech. This helps illiterate people who cannot read and/or write, hands busy people, transcribers of legal, medical, meeting (parliamentary for example) speeches. The need for the ASR technologies and the advancements in powerfull algorithms, such as deep learning, processing power with graphics processing unit (GPU) and availabil- ity of large amounts of data have encouraged research and development in ASR. As a result, a considerable amount of research on the development of ASR Systems (ASRS) has been conducted and lots of ASRSs have been developed. So far, however, ASRSs have only been developed for …

Content truncated to 500 words