Sarcasm Detection

This project implements sarcasm detecrion by CNN(Shuhui Wu) and Bert(Shuqi Gao)

Dataset

• We obtain dataset of Labelled Sarcastic Comments from Kaggle: https://www.kaggle.com/sherinclaudia/sarcastic-comments-on-reddit
• In CNN implementation, we use pretrained GloVe word embedding: https://nlp.stanford.edu/projects/glove/
• In BERT, we use google's pretrained bert by two different package:https://github.com/google-research/bert, https://github.com/huggingface/transformers

Approach

• CNN: https://www.aclweb.org/anthology/N18-2018.pdf       notebook:  Final_proj_CNN.ipynb
• BERT: papper: https://arxiv.org/pdf/1810.04805v2.pdf
  followed:  https://colab.research.google.com/github/tensorflow/tpu/blob/master/tools/colab/bert_finetuning_with_cloud_tpus.ipynb
  notebook: try_bert_1.ipynb and try-bert-2.ipynb