XPath Agent

XPath Agent is an advanced language model-based crawler designed to efficiently extract information from massive html files.

How to Reproduce

To reproduce the results presented in the associated paper, follow the steps below using the provided command-line interface script.

Prerequisites

Ensure you have the necessary dependencies installed.

# create a new python environment via conda
conda create -n feilian python=3.10
# install dependencies via pip
python -m pip install -r requirements.txt

Configure environment variables. For example, if you are using OpenAI API, ensure following environment variables are setting correctly:

OPENAI_API_KEY=sk-xxx

Download SWDE & SWDE_Extended to data folder, and unzip.

• SWDE: https://academictorrents.com/details/411576c7e80787e4b40452360f5f24acba9b5159
• SWDE_Extended: https://github.com/cdlockard/expanded_swde

Running the Experiment

Use the following command to run the experiment:

python scripts/experiment_cli.py --dataset <DATASET> --ie_model <IE_MODEL> --program_model <PROGRAM_MODEL> --name <EXPERIMENT_NAME>

Command-Line Arguments

• --dataset: Specifies the dataset to use. Options are SWDE or SWDE_Extended.
• --data_dir: The directory where the dataset is stored. Default is data/swde.
• --ie_model: The information extraction model to be used.
• --program_model: The program model to be used.
• --ablation: Optional. Specifies the ablation study option. Choices are
• --eval_sample_size: The number of samples to evaluate. Default is 32.
• --output: The directory where the results will be saved. Default is .tmp.
• --env: The environment file to load variables from. Default is .env.
• --seed: The random seed for reproducibility. Default is 0.
• name: A unique identifier for the experiment.