Yunus Can Bilge

I'm an Assistant Professor of Computer Science at Hacettepe University. I have received a PhD degree in Computer Engineering from Hacettepe University. My PhD focused on learning sign languages with limited supervision and semantic representations.

Email  /  Scholar  /  X  /  Github  /  News  /  Research  /  Selected Publications  /  Teaching  /  Patents  /  Blog

News

• [January 2026] New paper at Computer Science Review: An overview of sign language processing from natural language processing perspective
• [June 2025] New paper at Knowledge Based Systems: Modest: A dataset for multi domain scientific title generation
• [June 2025] joined Hacettepe University as an assistant professor of computer science.

Research

I'm interested in:

• machine learning (data-efficient machine learning with minimal supervision; zero-shot, few-shot, self-supervised learning)
• multimodal machine learning (integrate and reason over multiple data modalities—such as text, vision, and audio)
• production-ready machine learning systems (deployment and scalability in real-world settings)
• computer vision (image and video understanding)

Previously, I worked on sign language processing and machine learning systems in video surveillance settings.

Selected Publications

The paper provides an NLP-centric overview of sign language processing, identifying core and missing tasks, analyzing shared challenges with NLP, and proposing linguistically grounded research directions.

Introduces a multi-domain and multilingual dataset for scientific title generation in English and Turkish, evaluates LLMs across learning settings, and highlights the value of abstracts and domain-aware modeling.

Addresses zero-shot sign language recognition via textual/attribute semantic class representations, with benchmarks and a model combining visual and semantic features for unseen sign classes.

Embedding-based framework leveraging spatio-temporal visual and hand landmark features, with multilingual benchmarks for recognizing novel signs with limited examples.

Introduces WildestFaces and a clean-to-violent domain transfer framework for recognizing identities in violent videos.

Introduces ZSSLR using textual descriptions and a spatiotemporal framework combining video and text embeddings.