Fixy-TR/DeepChecker
📦 Open Source ProjectFixy-TR
A deep learning-based library for accurate detection and correction of Turkish clitics like da, de, ki, and mi.
DeepChecker serves as a dedicated NLP tool for the Turkish language, focusing on the frequent and often confusing usage of clitics. In Turkish, the distinction between attached and detached particles (like 'de/da' or 'ki') is a common source of grammatical errors. This library addresses this by employing deep learning models, specifically utilizing BERT-based architectures and Keras for training and inference. The project is built to integrate into larger data science pipelines, allowing developers to clean text data or build automated proofreading systems. By moving away from static rule-based scripts, DeepChecker adapts to contextual nuances in text, making it a valuable asset for Turkish-language content analysis, sentiment analysis preprocessing, and automated text correction applications.
💡Highlights
- ├─BERT-based clitic detection
- ├─Supports da, de, ki, mi particles
- └─Python-based Keras implementation
🎯For
- ├─NLP Researchers
- ├─Turkish Language Developers
- └─Data Scientists