Abstract:
"
In the twenty-first century, people have become more digitalized than before.
Microblogging platforms are the primary source of the communication network.
Microblogging platforms give users the freedom to express their feelings. After
coronavirus streaks, there is an increment in microblogging platforms because this is
the only way of connecting to people. Since coronavirus started in Wuhan, people
thought that china had created the virus and started discriminating against Asians.
There is a rise in online discrimination due to specific incidents that happen in 2020.
The prevention of hate content is essential before it could get uncontrollable. This
research study forces on classifying hate and counter hate speech towards Asians in
microblogging platforms.
Anti-Asian Detection Platform helps develops and researchers to classify hate and
counter hate speech with the help of transformation learning. Classifier considers not
only text content. It also considers emojis, hashtags before classifying. The
classification model is built customizing the BERT model and with the help of a new
fine-tuning strategy to the BERT model. Anti-Asian Detection Platforms classification
model uses multilayer perceptron with BERT model.
The final model is identified after experimenting with available methods. A new
hypothesis is introduced for classifying hate and counter hate speech towards Asians
in a pandemic situation. A web service is developed to help developers to integrate a
model that can identify how user behaviours. The annotated dataset was produced with
new hashtags that can be used to fetch hate and counter hate content. Anti-Asian
Detection Platform has outperformed similar systems. The classification model was
evaluated with recall, precision and F1 score. The following are the results for the
categories, hate F1 score is 0.62, the counter hate F1 score is 0.76, and Neutral F1
score is 0.73. Anti-Asian Detection is further evaluated with the help of domain and
technical experts. "
