
EMIR v1.0 was created on July 21, 2021". This project aims to continuously grow the Ethiopian music dataset to encourage other researchers to explore Ethiopian music.
There are three types of recording in EMIR: Firstly, Ethiopian Orthodox Tewahedo chants form part of a religious tradition dating back to the time of Saint Yared, secondly traditional Ethiopian Azmaris (songs), and thirdly modern secular Ethiopian music. The Orthodox chants were collected from online sources such as YouTube and DireTube. Some Azmaris were specially recorded in Addis Ababa by an ethnomusicologist specializing in Azmari houses; these are traditional venues where Azmaris are studied and performed. Firstly, five typical Azmari houses were selected for the study. Secondly, these were visited on multiple occasions. Each time, a singer was asked whether they would record an Azmari of their choice in a specified Kiñit. It went ahead if the singer knew and Azmari in that Kiñit and agreed to the recording. Otherwise, another singer was asked. In this way, a collection of Azmaris in the different Kiñits was built over several visits to each house. Further, Azmaris were collected from online sources such as YouTube. Finally, secular music was collected from online sources.
Every recording was independently reviewed by all five judges. Judges were responsible for the quality control of the dataset. Each Judge listened independently to all the recordings. For each one, they either assigned it to one of the four Kiñits or rejected it as not falling into any one of them. If three or more judges assigned a recording to the same category, EMIR accepted it. Otherwise, it was rejected. We aimed to collect 1,000 recordings. However, because of the above selection process, many of these were rejected, resulting in 600: 162 Tizita, 144 Bati, 147 Ambassel, and 147 Anchihoye
Four well-known Ethiopian Kiñits (pentatonic scales), Tizita, Bati, Ambassel, and Anchihoye
The tracks are all 16 kHz Mono 16-bit audio files in .wav format. Each file was labeled in the form `Bati1.Wav'. The first part of the name indicates the Kiñit (Tizita, Bati, Ambassel, or Anchihoye); the second part indicates the recording number within that Kiñit (1, 2, 3...). Subsequently, recordings were stored in four different folders in the dataset.