Is Less More? Quality, Quantity and Context in Idiom Processing with Natural Language Models

Adjacent code related to the paper prepared for Joint Workshop on Multiword Expressions and Universal Dependencies (MWE-UD 2024), 25th May, 2024. Collated with LREC-COLING 2024 in Torino, Italy.

Originaly prepared as part of a Masters dissertation project on COM4520 at The University of Sheffield, with the name of "Is Less More? Idiom Detection with Generalised Natural Language Models".

Created by: Agne Knietaite, Adam Allsebrook, Anton Minkov, Adam Tomaszewski, Norbert Slinko, Richard Johnson.

Supervised by: Aline Villavicencio.

Link to the paper: https://doi.org/10.48550/arXiv.2405.08497

Overview of the Repository

The code is divided into two sections, Model and Dataset related, following the structure of the adjacent report. The subdirectories are almost always self-contained and have their own README files for clarity.

Model Related

• src: the base model code for fine-tune and pre-train scenarios. Equipped with model performance evaluation and visualisation techniques;
• Jupiter Notebooks: Jupiter notebooks, which allow to run the models as Jupiter Notebooks on Google Collab;
• HPC Fine Tunning: Information on how to schedule model training as jobs on other machines;
• Paragraph External Context: Gathering external context for the external knowledge enhanced model described in the paper

Dataset Related

• Data Augmentation: Everything related to the data augmentation, used in the report;
• Data Generation Parser: Main parser, used for scrapping, adapting and constructing the datasets;
• Datasets: Base datasets used throughout the project;
• Script For Scraping Text File & Web Scrapper & Crawler: initial and later improved versions of data scraping and dataset forming tools.