The Turing Game
A multiplayer human–AI conversation game, followed by experiments on whether language models and detectors can tell human and bot participants apart.
Lecturer · Scholar
Language,
studied closely.
I work on language technologies for Turkish, from language models and evaluation benchmarks to the linguistic questions that generic multilingual systems tend to miss. I also teach how contemporary language models are built, adapted, and questioned.
Research
Research
My recent work focuses on Turkish language models and evaluation: building open models such as TURNA and TabiBERT, and designing benchmarks that test what these systems can actually do in Turkish. Earlier work examined named entity recognition and morphology in Turkish and other morphologically rich languages.
Google Scholar ↗Selected publications
2025
TabiBERT: A Large-Scale ModernBERT Foundation Model and Unified Benchmarking Framework for Turkish ↗
Melikşah Türker, Asude Ebrar Kızıloğlu, Onur Güngör, Susan Üsküdarlı.
arXiv preprint.
2024
TURNA: A Turkish Encoder-Decoder Language Model for Enhanced Understanding and Generation ↗
Gökçe Uludoğan, Zeynep Yirmibeşoğlu Balal, Furkan Akkurt, Melikşah Türker, Onur Güngör, Susan Üsküdarlı.
Findings of the Association for Computational Linguistics: ACL 2024.
2021
Neural Named Entity Recognition for Morphologically Rich Languages
PhD thesis, Boğaziçi University.
2019
The effect of morphology in named entity recognition with sequence tagging
Natural Language Engineering.
2018
Improving Named Entity Recognition by Jointly Learning to Disambiguate Morphological Tags
COLING.
2017
Morphological embeddings for named entity recognition in morphologically rich languages
2017
Linguistic features in Turkish word representations
2010
Morphological annotation of a corpus with a collaborative multiplayer game
Teaching
Teaching
At Boğaziçi University, I teach CMPE 586, Transformer-based Pre-training Methods in Natural Language Processing. The course moves from language modelling, attention, Transformers, and BERT to decoder-only models, training smaller language models, retrieval-augmented generation, multimodality, on-device systems, model merging, and agents. Students read research, work with contemporary tools, and develop a project around a linguistic problem or an application of language models.
Course outline ↗Advising
Student projects
Some of the most interesting teaching continues outside the classroom: students turn open questions into working systems, experiments, and public resources.
On-device renarration
A Chrome extension prototype that adapts web text to configurable reader profiles, exploring how renarration can happen locally in the browser.
Visual analysis of TBMM transcripts ↗
A visual and semantic exploration tool for patterns in Turkish parliamentary proceedings.
Background
Education
PhD (2021), MS (2009), and BS (2006) in Computer Engineering, Boğaziçi University. Part-time faculty at Boğaziçi University and Data Science Manager at Udemy.