File size: 758 Bytes
d8d6c9f
 
 
 
 
 
 
62f5b61
d8d6c9f
 
 
 
 
9d994ae
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
---
language: en
tags:
- transfer-learning
- bert
- hatespeech
- covid19
license: "mit"
datasets:
- COVID-HATE
metrics:
- f1-score
---
Since the start of the COVID-19 pandemic, there has been a widespread increase in the amount of hate-speech being propagated online against the Asian community. This project builds upon and explores the work of He et al. Their COVID-HATE dataset contains 206 million tweets focused around anti-Asian hate speech. Using tweet data from before the COVID-19 pandemic, as well as the COVID-HATE dataset from He et al, we performed transfer learning. We tested several different models, including BERT, RoBERTa, LSTM, and BERT-CNN.
Some of these models hindered the performance of He et al’s model, while others improved it.