Source code for Target-phrase Zero-shot Stance Detection: Where Do We Stand?

Requirements

python==3.8.10

more: requirements.txt

Our experiments

Our experiments scripts are in folder experiments-scripts. Details:

• Sem2016T6 BERT-base experiments: semeval-bert-base-verb-vs-ff-vs-cls-1.py, allaway-seval-bert-baseline-1.py
• Sem2016T6 BERT-large experiments: semeval-bert-large-cls-1.py
• Sem2016T6 ROBERTA-base experiments: semeval-roberta-base-verb-vs-ff-vs-cls-1.py
• Sem2016T6 ROBERTA-large experiments: semeval-roberta-large-verb-ff-cls-prompts-1.py
• VAST BERT-base experiments: vast-bert-base-1.py, vast-bert-base-2.py, vast-bert-base-3.py
• VAST BERT-base prompts experimentsvast-bert-base-ff-vs-verb-1.py
• VAST BERT-large experiments: vast-bert-large-1.py
• VAST RoBERTa-base experiments: vast-roberta-base-1.py
• VAST RoBERTa-large experiments: vast-roberta-large-1.py

Reproduced experiments

Code changes required for reproducing experiments are in folders: JointCL, VTCG, WS-BERT, BS-RGCN.

LLMs experiments

All prompts used in our experiments with LLMs are in llm-prompts.md file.

Datasets files

Datasets should be put in data directory with following filenames:

• Sem2016T6: se_train.csv, se_test.csv
• VAST: vast_train.csv, vast_dev.csv, vast_test.csv. datasets-processing/vast_topic_prep.py should be run before running experiments.