Open topics for Master’s thesis



Interpretation of Multimodal Deep Learning Models: provide interpretable outputs for neural model trained on multimodal (image-text, video) data Relevant publications

Visual Framing Analysis of Social Media Posts for Detection of Misinformation: analyse how visual information is depicted in social media posts and use it for misinformation identification. Relevant publications

Multimodal Hate Speech Detection in Videos: detecting visual and verbal hate speech in Social Media posts, videos etc. Relevant publications

Quality Assestment of Educational Videos: analyse educational videos from MOOC to detect whether they include low quality video in terms of audio and visual appeal. Relevant publications

Classification of Age Appropriateness of Movie Trailers: build a model to analyse movie trailers to classify for different age groups such as “Under 13”, “Over 18” etc. Relevant publications

Text Summarization for Scientific Domain: build a model to that summarizes a set of scientific articles and creates a section such as “Related Work” in research publications. Relevant publications

Extract Dataset and Performance Measures from Scientific Publications: build a model to extract the names of datasets and performance measures such as F1-scores, Mean Average Precision, etc. from scientific publications. Relevant publications

Detecting Reuse of Scientific Facts in News Domain: build a model to detect how scientific facts are used in news domain, e.g. news articles, videos, social media posts, etc. Relevant publications

Detection of Visual Violence in Social Media: build a model to detect how violence is expressed in social media posts as images or videos Relevant publications

Detection of Misinformation in Videos from Social Media: build a model to detect misinformed facts that are present in social media videos Relevant publications

Required skills: Python, Familiarity with any Deep Learning library (Pytorch, Tensorflow), Natural Language Processing, Computer Vision