Aniemore (Artem Nikita Ilya EMOtion REcognition)

2024: Alibaba's use in testing a new AI training method

Alibaba began using the Aniemore library datacet (Artem Nikita Ilya EMOtion REcognition), developed by the Russian company Social Code, in testing a new method of training artificial intelligence. This became known in February 2024.

Alibaba, together with researchers from three Chinese universities, created an author's model for recognizing emotion2vec emotions. To compare performance and metrics, foreign colleagues included in the study and used the comparative dataset of emotional speech dialogues (RESD), prepared for the Aniemore open library, specified in the Social Code company.

Pexels

They noted that when developing AI, it is important to have data sets (datacets) to compare model performance with existing models. For this, well-known and proven datacets having a certain quality, description and content are taken.

Engineers at universities in China and Alibaba took RESD as one of the metrics on which they tested their development and determined the quality of its work. It's nice that for these purposes, they took advantage of our base, not including even such well-known datacets as Dusha (Sberbank) in the tests. This is a victory and international recognition of our work, - said Artem Amentes, leader of the development of the Aniemore library.

By February 2024, the Russian Emotional Speech Dialogues dataset contains more than 3 thousand audio fragments from 200 different people, the model is able to recognize emotions in noisy audio files up to three seconds long. The datacet also contains real dialogues of high quality and different emotional colors. To prepare these materials, "Social Code" attracted professional actors to the work.^[1]

Notes