Results and code

This repository contains the results and code from the PhD thesis Extraction of Quantitative Grammatical Rules from Syntactic Treebanks.

Results

All results are in the results directory. The JSON files contain the scope, the conclusion and the extracted rules for each experience of the thesis.

Code

The main code for rule extraction is in src. Within this directory:

• grex2: the rule extraction scripts, including those for the decision tree, sparse logistic regression, and the RuleFit implementation.

The work is built upon a modified version of Grex2. The original Grex2 project is maintained at https://github.com/FilippoC/grex2. The code was primarily developed by Caio Corro.

The official documentation for Grex2 (currently under development) can be found at https://grex.grew.fr.

If you use this software, please cite our paper:

@inproceedings{herrera2024grex,
    title = "Sparse Logistic Regression with High-order Features for Automatic Grammar Rule Extraction from Treebanks",
    author = "Herrera, Santiago and Corro, Caio and Kahane, Sylvain",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation",
    month = may,
    year = "2024",
    address = "Torino, Italia",
    url = "https://arxiv.org/abs/2403.17534",
}

• univariate: script for compute univariate measures over features.
• evaluation: evaluations and global measure scripts.

Experiments

This directory includes bash scripts for executing the tasks of each experiment. It also contains other scripts or notebooks specific to each experiment.

Setup

Clone the repository and then install the project with pip install -e .

I recommend installing the project in a virtual environment, e.g., python -m venv .venv

Some paths need to be adjusted in order to run the experiments.