The democratization of deep learning in TASS 2017

Manuel C. Díaz-Galiano, Eugenio Martínez-Cámara, M. Ángel García Cumbreras, Manuel García Vega, Julio Villena Román


TASS 2017 has brought advances in the state-of-the-art in Sentiment Analysis in Spanish, because most of the systems submitted in 2017 were grounded on Deep Learning methods. Moreover, a new corpus of tweets written in Spanish was released, which is called InterTASS. The corpus is composed of tweets manually annotated at document level. The analysis of the results with InterTASS shows that the main challenge is the classification of tweets with a neutral opinion and those ones that do not express any opinion. Likewise, the organization exposed the project of extending InterTASS with tweets written in different versions of Spanish.

Texto completo: