| | --- |
| | license: cc-by-nc-3.0 |
| | --- |
| | |
| | PragS1: Pragmatic Masked Language Modeling with Hashtag_end dataset followed by Emoji-Based Surrogate Fine-Tuning |
| | |
| | You can load this model and use for downstream fine-tuning. For example: |
| | ```python |
| | from transformers import AutoTokenizer, AutoModelForSequenceClassification |
| | |
| | tokenizer = AutoTokenizer.from_pretrained('UBC-NLP/prags1', use_fast = True) |
| | model = AutoModelForSequenceClassification.from_pretrained('UBC-NLP/prags1',num_labels=lable_size) |
| | ``` |
| | |
| | |
| | More details are in our paper: |
| | ``` |
| | @inproceedings{zhang-abdul-mageed-2022-improving, |
| | title = "Improving Social Meaning Detection with Pragmatic Masking and Surrogate Fine-Tuning", |
| | author = "Zhang, Chiyu and |
| | Abdul-Mageed, Muhammad", |
| | booktitle = "Proceedings of the 12th Workshop on Computational Approaches to Subjectivity, Sentiment {\&} Social Media Analysis", |
| | month = may, |
| | year = "2022", |
| | address = "Dublin, Ireland", |
| | publisher = "Association for Computational Linguistics", |
| | url = "https://aclanthology.org/2022.wassa-1.14", |
| | pages = "141--156", |
| | } |
| | ``` |
