README.md

Code for "Ranking Based Multi-Label Classification for Sentiment Analysis" LKE 2019(7th International Symposium on Language & Knowledge Engineering).

Steps to run

1. Download datasets

   • Ren_CECps dataset, download it by contacting [email protected].
   • SemEval2007.
   • Sina Social News dataset, download it by contacting [email protected].

   Unzip these datasets in data folder and use the parser.py to convert them into .json format.

   mkdir data && cd data
   # unzip the datasets in `data`
   python3 parser.py

2. Download pretrained model

   mkdir models && cd models
   mkdir bert-base-chinese && cd bert-base-chinese
   wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-chinese.tar.gz
   tar -zxvf bert-base-chinese.tar.gz
   wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-chinese-vocab.txt
   cd ..
   mkdir bert-base-uncased && cd bert-base-uncased
   wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased.tar.gz
   tar -zxvf bert-base-uncased.tar.gz
   wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txt

3. Train the model

   python3 bert_classifier.py

References

• Pytorch-transformers

• Sentence emotion analysis and recognition based on emotion words using ren-cecps

• Relevant Emotion Ranking from Text Constrained with Emotion Relationships