Аннотация:This paper describes the SemEval2016 shared task on Aspect Based Sentiment Analysis (ABSA),a continuation of the respective tasks of 2014 and 2015. In its third year, the task
provided 19 training and 20 testing datasets for 8 languages and 7 domains, as well as a
common evaluation procedure. From these datasets, 25 were for sentence-level and 14 for
text-level ABSA; the latter was introduced for thefirsttimeasasubtaskinSemEval. Thetask
attracted 245 submissions from 29 teams.