Demo: SensiTrain: A Crowd Supported Platform to Understand Context and Improve Sensitivity in Online Communication

Pushwitha Krishnappa, Tahmid Ahmed, Otabek Abduraufov, Tathagata Mukherjee, Xiaoti Fan, Sriram Chellappan

Abstract: In this work we present SensiTrain - a data collection system designed with the goal of curating a dataset for understanding how differences in cultural backgrounds, sexual orientation, gender, race and national origin affect the way one perceives statements made on social media as being either benign or hurtful and sometimes even hateful and aggressive. SensiTrain displays social media posts to users and asks them to categorize the posts as being either benign or hurtful while providing a reason for such classification. We plan to use the data thus collected to build a public dataset repository and use the same to train a baseline system for identifying potentially hurtful statements before they are posted on social media platforms. The resulting system would be able to explain why a given statement could be perceived as hurtful by one or more groups thus making the inference more trustworthy to the users of the system. We hope that through this endeavor we will be able to mitigate unintentional harm caused due to lack of understanding of different cultures and backgrounds, thus making the world a kinder place.

External IDs:dblp:conf/dev/KrishnappaAAMFC24