Not all topics are equally “flammable” in terms of toxicity: a calm discussion of turtles or fishing less often fuels inappropriate toxic dialogues than a discussion on politics or sexual minorities. We will define a set of sensitive topics that can yield inappropriate and toxic messages and describe the methodology of collecting and labelling a dataset for appropriateness. While toxicity in user-generated data is well-studied, we aim to define a more fine-grained notion of inappropriateness. The core of inappropriateness is that it can harm the reputation of a speaker. This is different from toxicity in two respects: (i) inappropriateness is topic-related, and (ii) inappropriate message is not toxic but still unacceptable. We will collect and release two datasets in Russian: a topic-labelled dataset and an appropriateness-labelled dataset. We will also release pre-trained classification models trained on this data. The talk is based on the recent publication at the BSNLP workshop at EACL-2021 conference: https://www.aclweb.org/anthology/2021.bsnlp-1.4/


Nikolay Babakov
Research engineer
Don't miss
Mon Aug 02 2021 12:21:14 GMT+0300 (Moscow Standard Time)