Marathi_Movie_Script_Corpus_by_Pratik_Kalamkar

DescriptionMarathi_Movie_Script_Corpus_by_Pratik_Kalamkar is a curated collection of 100 Marathi movie script-like text files developed to support Natural Language Processing (NLP), Computational Linguistics, and Artificial Intelligence research.

Unlike English and Hindi, publicly available Marathi movie scripts are extremely difficult to obtain. During the creation of this corpus, it was found that very few complete Marathi movie scripts were available online. To address this limitation, subtitle resources from 100 Marathi films were collected from multiple publicly accessible sources and transformed into machine-readable script-like documents suitable for computational analysis.

The original subtitle files contained timestamps, sequence numbers, formatting tags, metadata, and fragmented dialogue spread across multiple lines. Extensive preprocessing was therefore required to convert these resources into a standardized corpus.

The preprocessing workflow included:

• Collection of subtitle resources from multiple public sources• Translation and language normalization where required• Removal of timestamps and subtitle metadata• Removal of formatting artifacts and sequence numbers• Merging fragmented subtitle lines into continuous text• UTF-8 encoding standardization• Regular-expression-based cleaning and normalization• Formatting standardization across files• Manual quality assurance and verification

The resulting corpus consists of script-like textual documents containing dialogues, sound cues, non-verbal expressions, and narrative content that closely resemble screenplay data while retaining the linguistic richness of movie conversations.

The final corpus contains:

• 100 Marathi movie script-like files• 153,565 lines of text• 851,377 words• UTF-8 encoded Devanagari text

SignificanceThis dataset provides one of the earliest publicly documented large-scale Marathi movie corpora specifically prepared for NLP research. The corpus offers researchers access to a substantial Marathi language resource suitable for developing and evaluating computational linguistic tools and machine learning models.

Potential applications include:

• Part-of-Speech Tagging• Morphological Analysis• Named Entity Recognition• Sentiment Analysis• Text Classification• Language Modeling• Machine Translation• Information Extraction• Computational Linguistics Research• Low-Resource Language NLP

NoveltyThe novelty of this dataset lies in transforming fragmented subtitle resources into a unified machine-readable Marathi movie corpus. At the time of creation, no publicly available large-scale Marathi movie-script corpus suitable for NLP experimentation and benchmarking was available. This dataset bridges that gap by consolidating, cleaning, normalizing, and standardizing data from multiple heterogeneous sources into a reusable research resource.

Citation

If you use this dataset in academic research, please cite:

Kalamkar, P. N., Peddi, P., & Sharma, Y. K. (2026).

A Comparative Analysis of PoS Tagging Tools for Hindi and Marathi.

International Journal of Informatics and Communication Technology (IJ-ICT), 15(1), 120–137.

DOI: https://doi.org/10.11591/ijict.v15i1.pp120-137

BibTeX:

@article{kalamkar2026postagging,author = {Kalamkar, Pratik N. and Peddi, Prasadu and Sharma, Yogesh K.},title = {A Comparative Analysis of PoS Tagging Tools for Hindi and Marathi},journal = {International Journal of Informatics and Communication Technology},volume = {15},number = {1},pages = {120--137},year = {2026},doi = {10.11591/ijict.v15i1.pp120-137}}