SemEval-2026 Task 11 - HABIB_TAZ

This repository contains the training scripts, data generation pipelines, and relevant documentation for our team's participation in SemEval-2026 Task 11.

Repository Structure

The repository is organized into the following main directories:

• data_generation/: Contains scripts, notebooks, and pipelines to generate both simple and complex synthetic syllogisms across multiple languages.
  • st1_st3/: Data generation scripts targeting Subtask 1 and Subtask 3.
  • st2_st4/: Data generation scripts and Jupyter Notebooks targeting Subtask 2 and Subtask 4, including generation in multiple languages (e.g., French, German, Spanish, Bengali, Swahili, etc.).
• training_scripts/: Contains the model training environments for the different subtasks.
  • st1.ipynb
  • st2.ipynb
  • st3.ipynb
  • st4.py
• paper/: Contains related documentation or paper materials for the submission.

Data Generation Guidelines

Inside the data_generation/st1_st3 and data_generation/st2_st4 directories, you will find PDF guideline documents (ST3_Data_Augmentation_Pipeline.pdf and ST4_Data_Augmentation_Pipeline.pdf).

Important Note: While these PDFs explicitly provide the pipelines and structured guidelines to generate data for Subtask 3 (ST3) and Subtask 4 (ST4), the exact same procedures can be trivially modified to generate data for Subtask 1 (ST1) and Subtask 2 (ST2) respectively. The primary difference is simply generating the syllogisms for a single language instead of multilingual generation.

Setup Requirements

Before running the scripts, please set up your Python environment:

1. Create and activate a Python virtual environment:

   python3 -m venv myenv
   source myenv/bin/activate

2. Install the necessary dependencies:

   pip install -r requirements.txt

Execution

After setup, refer to the individual scripts in data_generation/ and training_scripts/ for run instructions specific to each subtask.