The SEPLN International Congress is the key event on language technologies organized by the Spanish Society of Natural Language Processing . Aina’s technologies have been recognized during this year’s edition of SEPLN, held in Valladolid from September 24 to 27. Among the presentations, the study led by Marc Casals , et. al, in which the recognition of emotions in the voice is studied through the attention mechanism. In fact, the proposal of the Aina Project, developed by the Language Technologies Unit of the Barcelona Supercomputing Center (BSC-CNS) has won first place in one of the competitions held at the IberLEF event during the SEPLN , the EmoSpeech2024.
“It is a solution that brings some of the most important promises such as the monitoring of patients’ emotions or the improvement of customer service systems closer.”
The field of affective computing is one that is in a higher state of growth. The detection, recognition and interpretation of human emotions is sought through algorithms and devices. In the framework of this study, a multimodal system is presented through which text and voice components are integrated to be able to carry out a process of classification of emotions given a voice message. It is a solution that approaches some of the most important promises of emotion recognition, such as monitoring the emotions of patients or improving customer service systems.
Marc Casals and Javier Hernando, researchers at the Barcelona Supercomputing Center (BSC-CNS).
However, it is a proposal that integrates “inputs” of text and voice. A fact that links it directly to the developments in the different ones, in which Aina’s team also works. In fact, through a collaboration with Pompeu Fabra University researcher Núria Bel and Valle Ruiz from the BSC-CNS, a corpus of linguistic acceptability data has been presented. It is an innovative development that makes available key resources for the training of models that incorporate the parameters of inclusion and linguistic accessibility. All the Aina technologies that have been recognized during this year’s edition of the SEPLN and presented by Aina follow the line of work marked by inclusion and diversity in languages. Through a wider representation of the linguistic realities in the data corpora, a more accurate and precise performance of the language models is achieved. Resource excellence and quality is critical to generating AI resources that are competitive and adoptable by all actors.
