Transgender Community Sentiment Analysis from Social Media Data: A Natural Language Processing Approach

Abstract

Yuqiao Liu

There are huge disparities in mental health outcomes among transgender people compared to the general population. Interpreting social media data posted by transgender people may help people better understand the emotions of these sexual minorities and intervene early. In this study, 300 social media comments posted by trans people were manually classified into negative, positive and neutral sentiment. Using 5 machine learning algorithms and 2 deep neural networks to build a sentiment analysis classifier based on labeled data. The results show that the annotations are reliable, with Cohen's Kappa scores exceeding 0.8 in all three categories. The LSTM model yielded the best accuracy over 0.85 and an AUC of 0.876. Next steps will focus on using advanced natural language processing algorithms on larger annotated datasets.

PDF