Getting Started

S2IT runs dependency parsing, builds train/dev data, trains the model, performs extraction and classification, then evaluates results.

Prerequisites

• Python environment:
  • Install dependencies: pip install -r requirements.txt
• Stanza model files:
  • ACOS-main/preprocess_dependency.py uses stanza
  • Ensure English models are available at stanza_resources/stanza-en/

One-Click Run

Run run.sh (default dataset=restaurant):

bash run.sh

The script does the following:

export CUDA_VISIBLE_DEVICES=

export model_name_or_path=Qwen/Qwen2.5-7B-Instruct
export template=qwen
python ACOS-main/preprocess_dependency.py
python ACOS-main/preprocess_training.py
python ACOS-main/preprocess_dev.py

export dataset=restaurant
bash train.sh
bash extraction.sh
python ACOS-main/pre_classification.py
bash classification.sh
python evaluate.py