Onan, Aytuğ

Profile URL
https://hdl.handle.net/11147/18906
Name Variants
Onan, Aytug
Onan, Aytuʇ
Job Title
Prof. Dr.
Email Address
aytugonan@iyte.edu.tr
Main Affiliation
Status
Current Staff
Website
https://aytugonan.github.io/
ORCID ID
0000-0002-9434-5880
Scopus Author ID
55201862300
Turkish CoHE Profile ID
172114
Google Scholar ID
cdUqOj0AAAAJ
WoS Researcher ID
L-4613-2018
Files
aytug_onan.jpg (12.34 KB)

Sustainable Development Goals

SDG data is not available
Scopus data could not be loaded because of an error. Please refresh the page or try again later.
WoS data could not be loaded because of an error. Please refresh the page or try again later.

Publication Collaboration

Affiliation Name Count
Izmir Kâtip Çelebi University 108
Manisa Celal Bayar University 53
Ege University 14
Necmettin Erbakan University 10
Izmir Institute of Technology 9
1 / 6
Data obtained from OpenAlex
Scholarly Output

4

Articles

4

Views / Downloads

22/0

Supervised MSc Theses

0

Supervised PhD Theses

0

WoS Citation Count

0

Scopus Citation Count

0

Patents

0

Projects

0

WoS Citations per Publication

0.00

Scopus Citations per Publication

0.00

Open Access Source

1

Supervised Theses

0

JournalCount
IEEE Access1
Information Sciences1
Interactive Learning Environments1
Quality and Quantity1
Current Page: 1 / 1

Scopus Quartile Distribution

Competency Cloud

GCRIS Competency Cloud
Author of Publication: search.filters.isAuthorOfPublication.fc142cb1-08fe-4116-aedf-609f5769e24d

Scholarly Output Search Results

Now showing 1 - 4 of 4
  • Article
    A Multidimensional Comparative Analysis of Human Expert vs. AI-Driven Feedback Approaches on Learner-Centered and Collaborative Groups
    (Routledge, 2026) Onan, Aytuğ; Onan, A.; 01. Izmir Institute of Technology
    The aim of this study is to examine the multidimensional effects of AI-based feedback in learner-centered and collaborative learning environments among university students. The study employed a five-group experimental design: two individual learning groups receiving either AI-based feedback(G1) or human expert feedback(G2), two collaborative learning groups receiving either AI(G3) or human expert feedback(G4), and control group(G5). According to the research results, G4 showed the highest level of development in the areas of creative problem solving, internal-external motivation, and critical thinking. G1 was the group with the highest performance, particularly in terms of system interaction, completed activities, and assignments. In contrast, G2 showed the lowest results in terms of both cognitive development and learning analytics. AI-based feedback in collaborative learning environments provided the highest development in cognitive skills, while AI-based in individual work was more effective in increasing system participation. Factorial MANCOVA indicated significant interactions between learning environment and feedback type on posttest outcomes, with strongest effects on self-efficacy, intrinsic motivation, and flexibility. These results show that AI-based feedback has different effects in both individual and collaborative learning contexts. Qualitative thematic analysis highlighted themes of cognitive facilitation, creativity enhancement, feedback quality perceptions, and environment preferences. © 2026 Informa UK Limited, trading as Taylor & Francis Group.
  • Article
    Interpretable Structural Modeling of MR Images Using Q-Bezier Curves: A Geometry-Aware Paradigm Beyond Deep Learning
    (Elsevier Science Inc, 2026) Onan, Aytuğ; Onan, Aytug; Turhan, Nezihe; Ozger, Zeynep Odemis; 01. Izmir Institute of Technology
    Magnetic resonance (MR) imaging plays a critical role in diagnostic workflows, yet its reliability is frequently compromised by scanner-dependent bias, contrast variability, and intensity drift. Although deep learning methods achieve high performance, they generally require extensive supervision and demonstrate limited robustness across diverse clinical settings. To address these challenges, we propose a transparent, geometry-aware framework for annotation-free MR enhancement based on q-B & eacute;zier curves. This model incorporates an adaptive deformation parameter q(x) that modulates local curvature, facilitating flexible adaptation to complex anatomical boundaries. The framework comprises three principal mechanisms: (i) adaptive q(x) for local responsiveness, (ii) monotone q-Bezier tone curves for intensity standardization, and (iii) Tikhonov-regularized optimization for smooth mapping. As a result, the operator remains interpretable, operates in linear time, and provides explicit control over smoothness. The proposed approach was validated across five public cohorts (BraTS, ACDC, PROMISE12, fastMRI, IXI), demonstrating significant improvements in image fidelity (SSIM, CNR, NIQE) and downstream segmentation accuracy (Dice, HD95) relative to variational filters and state-of-theart foundation models. Additionally, cross-vendor experiments confirm its robustness without the need for retraining. Collectively, these findings establish q-Bezier modeling as a principled, lightweight, and clinically interpretable alternative that complements deep learning by providing a geometry-aware pathway to robust MR representation.
  • Article
    AI-Powered Detection of Hate Speech Against Refugees in Turkish Social Media
    (Springer Science and Business Media B.V., 2025) Onan, Aytuğ; Bulut, V.; Onan, A.; 01. Izmir Institute of Technology
    Social media platforms can cause hate speech to spread rapidly, so it is important to address such content. The speed at which hate speech spreads on social media makes it impossible to obstruct such content manually. Artificial intelligence support can be a solution for this. Detecting hate speech with artificial intelligence requires determining which expressions are hate speech. In this research, a study was conducted specifically on hate speech against refugees. Considering that Turkiye is the country with the highest migration after the Syrian Civil War, and the hospitality to approximately 3.9 Syrian people, the study focused on the Turkish language. The research first aims to create a Turkiye dataset of social media posts. Posts containing hate speech were labeled using discourse analysis on this dataset. The next stage of the research is to detect Turkish hate speech against refugees with artificial intelligence. According to the research results, the BERTurk model trained with this dataset achieved an accuracy rate of 85% in the automatic detection of Turkish hate speech. In the current climate, hate speech can spread rapidly in society and can easily lead to violent acts. Therefore, taking the necessary measures against hate speech is crucial. This study is crucial for automatically detecting hate speech in Turkish. © The Author(s), under exclusive licence to Springer Nature B.V. 2025.
  • Article
    Toward Reliable Annotation in Low-Resource NLP: A Mixture of Agents Framework and Multi-LLM Benchmarking
    (IEEE-Inst Electrical Electronics Engineers Inc, 2025) Onan, Aytuğ; Nasution, Arbi Haza; Celikten, Tugba; 01. Izmir Institute of Technology
    This paper introduces the Mixture-of-Agents (MoA) framework, a structured approach for improving the reliability of large language model (LLM)-based text annotation in low-resource NLP contexts. MoA employs coordinated agent interactions to enhance agreement, interpretability, and robustness without manual supervision. Evaluations on Turkish classification benchmarks demonstrate that MoA achieves up to 10-point improvements in macro-F1 over single-model baselines and significantly increases inter-agent consistency. Additionally, three novel reliability metrics-Conflict Rate (CR), Ambiguity Resolution Success Rate (ARSR), and Refinement Correction Rate (RCR)-are proposed to quantify annotation stability and correction dynamics. The results indicate that multi-agent coordination can substantially improve label quality, offering a scalable pathway toward trustworthy annotation in low-resource and cross-domain applications. The framework is language-agnostic and adaptable to other low-resource contexts beyond Turkish, including morphologically rich or typologically diverse languages such as Indonesian, Urdu, and Swahili. These findings highlight the scalability of MoA as a generalizable solution for multilingual and cross-domain annotation.