Offensive Language Identification for Twitter using BERT

International Workshop on Semantic Evaluation 2020 - Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval)

========================

DESCRIPTION

In this project we are implementing an offensive Language Identification Model for Twitter Tweets using a pretrained BERT model: bert_uncased_L-12_H-768_A-12/1

We train our model on a Google Cloud Computing Services using GPUs. However we do not recommed using GCS as they are not able to provide stable available ressources for the Virtual machines. In order to avoid downtime we recommend using other cloud services.

---

This project has been implemented by partly using ressources and code of:

https://github.com/SimonHegelich/ADL-NLP

https://norvig.com/spell-correct.html

https://github.com/google-research/bert/blob/master/predicting_movie_reviews_with_bert_on_tf_hub.ipynb