Poster/Short paper presentation by L3S at the 24th ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW 21) — 23 – 27 October 2021
Authors: Aparup Khatua and Wolfgang Nejdl
Abstract: Prior studies, mostly from the social science domain, have observed that mental stress and struggles are high for refugees. Information science researchers have found that social media data can be insightful for probing psychological stress. However, none of the previous studies, to the best of our knowledge, investigated social media data to identify the voices of migrants and refugees and analyze their concerns. We have collected 0.15 million tweets, but only 2% of these tweets are the voices of migrants and refugees. In addition to non-refugee and non-migrant voices, we have classified their voices into three themes as follows: their generic views, initial struggles, and subsequent settlement in the host country. We have employed deep learning and transformer-based models for identifying these themes. Our best-performing transformer-based model has reported an accuracy of 75.89%. We have also identified some exciting avenues for future research.
For more information about the conference click here