rerebabu/Tinta

An Automated System for Generating and Simulating Human-Like Filipino Grammar Errors with Artificial Rules for Data Augmentation # Tinta ✍️ *An Automated System for Generating and Simulating Human-Like Filipino Grammar Errors with Artificial Rules for Data Augmentation* ## 📌 Overview **Tinta** is a Python-based rule-driven system designed to simulate **Filipino grammar errors** in otherwise correct sentences. By modeling common mistakes found in actual student writings, Tinta generates realistic erroneous data for tasks like: - Data augmentation for machine learning - Grammar checking tool training - Language education and research This system is based on grammatical patterns identified in [#], particularly from student-written Filipino translation activities. --- ## ⚙️ How It Works 1. **Input**: Clean sentences are read from `sentences.txt`. 2. **Operations Applied**: - **Insert**: Adds random Filipino function words - **Delete**: Removes random tokens - **Substitute**: Introduces realistic grammar errors based on defined rules - **Swap**: Reorders nearby tokens 3. **Error Types Simulated**: - Use of ligatures - Use of enclitics - Hyphenation removal - “Ng” vs “Nang” confusion - Morphophonemic changes - Word repetition 4. **Output**: - A set of grammatically incorrect sentences - Summary tables of error and operation distributions ## 👥 Authors This is a collaborated project in **Introduction to Artificial Intelligence**. - **Jomar J. Ronquillo** - **Niñalene G. Paguio** - **Kenneth Claire M. Tulang** - **Reynalyn R. Bawag** --- ## 📚 References - Dataset: BALARILA Project - Error Taxonomy: **Octaviano, M., Go, M. P., Borra, A., & Oco, N. (2016)**. *A corpus-based analysis of Filipino writing errors.* In *International Conference on Asian Language Processing (IALP)*.