Authorship and linguistic evidence
I am interested in how writing style can support authorship attribution and verification, and how genre and context affect what we can infer from a text.
Applied linguistics · Stylometry · Explainable AI
Muhammad
Ismail.
A linguist’s perspective
on authorship & AI.
I am a lecturer at the Higher Education Department, Punjab, Pakistan. I study how patterns in language can inform our understanding of authorship. My research connects applied linguistics with computational stylometry, interpretable machine learning, and the careful evaluation of AI-generated writing.
My background spans English and applied linguistics, college teaching, quantitative research, and building research software.
Read my background ↗01 / ResearchQuestions
Questions
behind the work.
Explainable NLP
My research interests include explanation stability, behavioural faithfulness, and calibration: whether an explanation stays consistent, reflects model behaviour, and supports a careful reading of its output.
Human and AI-generated writing
My MA thesis compared 5,046 human and AI-generated texts across six genres, using 49 stylometric measures and SPSS. This statistical study is distinct from my later 132-feature classifier.
02 / Selected work
GitHub ↗Research, made tangible.
01 / Software
Style Scalpel
A 132-feature LightGBM system for examining AI-authorship signals in academic writing. It brings individual linguistic features, paragraph assessments, and inspectable reports into a working web application and Windows tool.
Stylometric feature data
A publication-safe dataset representing 5,046 texts through 132 extracted stylometric features, available through OSF.
Explore the dataset ↗Feature maps
An interactive resource for exploring linguistic features. The visualisation is a separate companion to the classifier.
Open the feature maps ↗03 / WritingPublications &
Publications &
research outputs.
Projection of mass extermination in Dan Brown’s Inferno: An apocalyptic study
Muhammad Ismail & Shahzeb Shafi
Journal of Academic Research for Humanities, 3(4), 126–136.
Style Scalpel: A Forensic Stylometric Tool for Identification of AI-Generated Academic Texts
Muhammad Ismail
Stylometric Comparative Analysis of Human and AI-Generated Texts
Eötvös Loránd University, Budapest. A quantitative comparative study of human and AI-generated writing.
04 / BackgroundFrom language study
From language study
to computational research.
Education
MA in English Studies
Applied Linguistics focus
Eötvös Loránd University · Budapest, Hungary
BS (Hons.) in English
National University of Modern Languages
Islamabad, Pakistan
Stipendium Hungaricum Scholarship
Tempus Public Foundation, Hungary
MA scholarship
Experience & methods
Lecturer
Higher Education Department, Punjab, Pakistan
College-level English teaching, course materials, assessment, and student feedback.
Research toolkit
Python, pandas, NumPy, scikit-learn, LightGBM, spaCy, SHAP, and SPSS. Corpus analysis, linguistic feature engineering, model evaluation, and reproducible workflows.
Languages
Urdu and Punjabi · native
English · IELTS 7.5
Hungarian · B1 examination
05 / ContactLet’s talk
Let’s talk
language & research.
For research conversations, collaboration, or questions about my work.
m0333ism@gmail.com ↗