No Detail Found!

Author Profiling From Short Romanized Urdu Messages: A Preliminary Investigation Using Transfer Learning Models

Author(s)
Abid Ali Department of Computer Software Engineering, University of Engineering & Technology Mardan,,Pakistan
Muhammad Sohail Khan Department of Computer Software Engineering, University of Engineering & Technology Mardan,,Pakistan
Muhammad Amin Khan Department of Computer Science, CECOS University of IT and Emerging Sciences,Peshawar,Pakistan
Abstract
Author profiling, a crucial task in natural language processing, involves identifying various attributes of an author, such as gender and age, from text. This study examines how transfer learning models in the context of author profiling from Roman Urdu text. We conduct experiments employing prominent models such as ELECTRA , BERT, RoBERTa, XLNet, Distil Bert, Distil RoBERTa,. Our analysis reveals superior performance in gender prediction using BERT, attaining an accuracy of 0.74698, precision of 0.7505, recall of 0.7462, and F1 score of 0.7456. On the other hand, RoBERTa demonstrates remarkable proficiency in age prediction with an accuracy of 0.8221, precision of 0.8215, recall of 0.8221, and F1 score of 0.8215. These findings showcase the efectiveness of transfer learning models in author profiling tasks ofer insightful analysis for further research and applications in this domain.
Publication Details
Page(s) 62-72
Published Journal: VFAST Transactions on Software Engineering, Volume: 11, Issue: 3, Year: 2023

Request the full text from PASTIC

Keywords
Transfer Learning, BERT, ROBERTA, Author Profiling, Roman Urdu
0

Citations

0

Downloads

101

Views