EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

An Efficient Machine Learning Framework for Roman Urdu Sentiment Analysis Using Hybrid Word–Character TF-IDF Features

Abdul Hannan, Layeq Ali, Rao Muhammad Asif

Pakistan Journal of Scientific Research · 30.12.2025 · 5 (02) · 161-169

DOI: 10.57041/cjc1as73 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Pakistan Journal of Scientific Research, online erschienen am 30.12.2025. Verfasst von Abdul Hannan und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The problem this study focuses on is sentiment analysis on Roman Urdu text due to its low resource status coupled with high informality and inconsistencies associated with the language. Specifically, Roman Urdu lacks standardized spellings and exhibits significant use of code mixing with English. Therefore, the main goal of the study is developing a computationally affordable machine learning model for performing sentiment analysis on Roman Urdu text. The research suggests applying a hybrid strategy of feature extraction based on term frequency-inverse document frequency (TF-IDF) word-level and character-level vectors to overcome issues related to spelling inconsistency and capture semantic context simultaneously. For the purpose of building different classifiers, the data set with more than 22,000 Roman Urdu reviews was employed. Four classifiers were selected for analysis – logistic regression, linear support vector machine, naïve Bayes, and random forest. Besides, a hard voting ensemble classifier was built to utilize advantages of individual classifiers. The results revealed that logistic regression had the highest individual accuracy (78.40%) whereas the ensemble classifier had the best overall result (79.19%).

Zitieren

APA

Hannan, A., Ali, L., & Asif, R. M. (2025). An Efficient Machine Learning Framework for Roman Urdu Sentiment Analysis Using Hybrid Word–Character TF-IDF Features. Pakistan Journal of Scientific Research, 5(02), 161-169. https://doi.org/10.57041/cjc1as73

Vancouver

Hannan A, Ali L, Asif RM. An Efficient Machine Learning Framework for Roman Urdu Sentiment Analysis Using Hybrid Word–Character TF-IDF Features. Pakistan Journal of Scientific Research. 2025;5(02):161-169. doi:10.57041/cjc1as73

BibTeX
@article{Hannan2025,
  author = {Hannan, Abdul and Ali, Layeq and Asif, Rao Muhammad},
  title = {An Efficient Machine Learning Framework for Roman Urdu Sentiment Analysis Using Hybrid Word–Character TF-IDF Features},
  journal = {Pakistan Journal of Scientific Research},
  year = {2025},
  volume = {5},
  number = {02},
  pages = {161--169},
  doi = {10.57041/cjc1as73}
}
Kostenlos starten →