Automatic sentence simplification system for Arabic Script Punjabi

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Corpus Linguistics

Background:

• Creating aligned monolingual parallel datasets for specific dialects is challenging.
• The Shahmukhi dialect of Punjabi has received limited attention in NLP research.
• Language simplification aims to improve text clarity and readability.

Purpose of the Study:

• To introduce the Punjabi Simplification (PUSIM) corpus, a novel dataset for the Shahmukhi dialect.
• To develop and assess a hybrid approach for Punjabi language simplification.
• To facilitate research in automated simplification for under-resourced languages.

Main Methods:

• Development of the PUSIM corpus focusing on the Shahmukhi dialect.
• Application of a hybrid approach for sentence simplification, enhancing clarity and vocabulary.
• Quantitative evaluation using automated readability metrics and the SARI score.

Main Results:

• The PUSIM corpus was successfully created for the Shahmukhi dialect.
• The hybrid simplification approach yielded high-quality results, evidenced by a SARI score of 45.3.
• Automated readability assessments confirmed improved text quality.

Conclusions:

• The PUSIM corpus represents a significant contribution to Punjabi NLP, particularly for the Shahmukhi dialect.
• The developed simplification methodology is effective and adaptable.
• This dataset will foster further research in automated language simplification for Punjabi.