sentiment-analysis-with-WSD

This repository contains the multitask learning model proposed in Neurosymbolic sentiment analysis with dynamic word sense disambiguation.

Usage

To pretrain and test the lexical substitution model, put the desired pretrained language model in the alm_path, and download the required datasets to the corresponding folders in the data folder. Then run the following example script:

python pretrain.py --batch_size 20 --lr 1e-8 --alm_path "./ckpt/saved_ckpt/ALM.pt"

To train and test the sentiment analysis model, put the path of the trained lexical substitution model in the lex_path. Then run the following example script:

python run.py --batch_size 10 --lr 1e-6 --lex_path "./ckpt/saved_ckpt/lex_sub.pt"

Citation

If you use this knowledge base in your work, please cite the paper - Neurosymbolic sentiment analysis with dynamic word sense disambiguation with the following:

@inproceedings{zhang-etal-2023-neuro,
    title = "Neuro-Symbolic Sentiment Analysis with Dynamic Word Sense Disambiguation",
    author = "Zhang, Xulang  and
      Mao, Rui  and
      He, Kai  and
      Cambria, Erik",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-emnlp.587",
    doi = "10.18653/v1/2023.findings-emnlp.587",
    pages = "8772--8783",
}