AI in Speech & Text Processing

2026 Projects

Format: On Campus or Online
Number of Students: 6
Duration of Project: 1 July-15 July

Project Supervisor: Cenk Demiroğlu
Research Areas: AI in speech & text processing
Daily Supervisor: Utku Ozan Çay Ozan Çay

Project Description
This short project focuses on building AI systems for detecting signs of depression from speech data. Students will implement a lightweight pipeline using pre-trained models and basic machine learning techniques. The emphasis is on practical implementation rather than research.

About the Project
Depression can affect both how people speak (acoustics) and what they say (language). In this project, students will use pre-trained speech recognition and audio models to extract features, and then build a simple classifier. They will also explore combining audio and text using a multimodal LLM.

The project is designed to be completed within two weeks with guided steps and provided resources.

Project Objectives

Build an end-to-end pipeline for speech-based classification
Use pre-trained models (no training from scratch)
Compare audio-based vs text-based features
Understand basic evaluation metrics (accuracy, F1-score)
Gain intuition about multimodal AI systems

Project Tasks

Load a small provided dataset (audio + labels)
Preprocess audio (resample, trim if needed)
Use a pre-trained ASR model (e.g., Whisper) to obtain transcripts
Extract simple features:

Audio embeddings (or basic features like MFCCs)
Text features (e.g., TF-IDF or embeddings)

Train a simple classifier (e.g., logistic regression or small neural network)
Evaluate performance and compare:

Audio-only
Text-only
Combined features (optional)

Deliverables

Working Python code
Trained model and evaluation results
A presentation describing:

Approach
Results
Observations

Benefits for the Student

Quick exposure to real-world AI workflows
Hands-on experience with speech and language models
Understanding of multimodal learning concepts
Practical skills using modern AI tools (e.g., Hugging Face, PyTorch)

Requirements

Basic Python knowledge
Introductory machine learning concepts
No prior experience with speech processing required

Projects for 2026

Applications Have Started!

Özyeğin University Summer Research Internship project is open to high school students and the aim of the program is to increase their experience on how scientific research is conducted.

2026 Projects