Student project ideas (science)
If you are enrolled at the University of Amsterdam, the Amsterdam University College, or the University of Bologna, and you would like to do a project/thesis/capstone with me, these are a few topics I or members of my team are interested into.
If you like one of them, or you would like to propose something else, get in touch: g.colavizza@uva.nl (please include program, year and grades).
Participation in shared tasks is very encouraged. Students from previous years have participated, for example, in challenges such as SemEval 2020 unsupervised lexical semantic change (task 1) and CLEF HIPE 2020 (Named Entity Recognition and Linking on multilingual historical corpora).
AI for cultural heritage
There are a variety of topics under this general theme which are of interest, as heritage collections have been widely digitized over recent years. Themes include the use of active learning to annotate and retrieve information (this is related to human-AI interaction or hybrid AI), the use of transfer learning (for example language models of virual features), the automated enrichment of collections via information extraction (for example, named entity recognition and linking), and the application of explainable AI methods to automated tasks in cultural or creative workflows.
Several of these topics are developed within the context of the Creative Amsterdam initiative (CREATE) at UvA, and in collaboration with Dutch heritage organizations.
For internships and industry projects under this theme, see below or visit Odoma.
To know more: Fiorucci et al., Machine Learning for Cultural Heritage: A Survey, 2020.
Computational art history
The availability of digitized (or born-digital) art collections allows scholars to use modern machine learning techniques to study artistic influences, styles and visual patterns recurring over time.
To know more: Shen et al., Discovering Visual Patterns in Art Collections with Spatially-consistent Feature Learning, 2019.
The language of science
How scientists use natural language to communicate? How does their use differs across disciplines (say, mathematics and philosophy)? How is science communicated to the public? How was this done historically? This projects aims at measuring the relative difference or similarity in the use of natural language across scientific communities, and among the public/media. Work on contemporary data (e.g., social media, Wikipedia and journal articles) as well as historical data (e.g., historical newspapers and scientific publications) is possible.
To know more: Ramage et al., Mapping Three Decades of Intellectual Change in Academia, 2020.
NFT art valuation and investment
We intend to develop accurate automated art valuation models combining market, artwork and social media data. The student will get access to Nifty Value’s data in order to design, develop and test models for the prediction of art prices, eventually designing and assessing investment strategies for different risk profiles.
This project is ideal for a student with a strong interest in applying AI to finance and art, and is linked to the activities of a start-up. See Nifty Value for more.
Augmenting art discovery
We intend to develop a recommendation engine based on the visuals of an artwork (aesthetics) as well as its market history. The personalised recommendation of art is a difficult task to partially automate as art appreciation is very subjective. The goal of the project is to develop a system which can recommend art as an investment (for profit) or according to personal taste (for pleasure).
This project is ideal for a student with a strong interest in applying AI to information retrieval, finance and art, and is linked to the activities of a start-up. See Nifty Value for more.
Industry internships
These projects are suitable for students with a strong background in AI and an interest in cultural heritage or the creative industries. Projects might be compensated as part of contracts via the consulting company Odoma.