Research

Papers and degrees

Three peer-reviewed papers: two on Arabic speech — dialect identification, then pronunciation scoring, which became my M.Sc. thesis — and one on catching hallucinations by aligning what a model says against a knowledge graph.

TrustNLP 2024

FactAlign: Fact-Level Hallucination Detection and Classification Through Knowledge Graph Alignment

Mohamed Rashad, Ahmed Zahran, Abanoub Amin, Amr Abdelaal, Mohamed Al-Tantawy

Proceedings of the 4th Workshop on Trustworthy Natural Language Processing, pages 79–84

Recasts black-box hallucination detection as knowledge-graph alignment, which also separates intrinsic from extrinsic hallucinations: 0.889 F1 on detection (WikiBio GPT-3) and 0.825 F1 on type classification (XSum), with no fine-tuning and no repeated sampling.

IEEE Access

Fine-Tuning Self-Supervised Learning Models for End-to-End Pronunciation Scoring

Ahmed I. Zahran, Aly A. Fahmy, Khaled T. Wassif, Hanaa Bayomi

Volume 11, pages 112650–112663

E2E-R scores pronunciation at the phoneme level straight from the waveform — SSL fine-tuning for phoneme recognition, then a Siamese comparison of pronounced against canonical phoneme embeddings. It reaches 0.68 PCC on speechocean762, comparable to the state of the art at the time, without extra native speech data, feature engineering, or external forced alignment.

Code on GitHub ↗

WANLP 2019

A Character Level Convolutional BiLSTM for Arabic Dialect Identification

Mohamed Elaraby, Ahmed Zahran

Proceedings of the Fourth Arabic Natural Language Processing Workshop, pages 274–278

Dialect identification straight from characters, avoiding the tokenisation problem that Arabic dialects create. Ranked 4th at 61.54% F1-macro on the MADAR 2019 shared task for fine-grained Twitter user dialect identification.

Education

June 2024

M.Sc. in Computer Science

Cairo University, Faculty of Computers and Information

Thesis: “Enhancement of Mispronunciation Detection Using Deep Learning Techniques”.
Proposed E2E-R, an end-to-end pronunciation scoring architecture using fine-tuned SSL speech models, matching the state of the art at the time on substantially less data and compute.

May 2016

B.Sc. in Computer Engineering

Cairo University, Faculty of Engineering

Graduation project: Animtractor, a marker-less motion capture system that needs no depth camera. Won first place in the Innovation track of the Microsoft Imagine Cup 2016 national finals.
Prepared cloud computing coursework and summer training for CMP303B — Distributed Operating Systems.
Member of the Academic Committee in IEEE’s Cairo University Student Branch.