Categories
Research Visit

The experience of a master’s student with TRIFECTA

By Katrina Slebos Perez

Before I started my master’s thesis, I knew I wanted to work on an interdisciplinary project. I study Artificial Intelligence but I’ve always been very interested in history and the humanities. Hence, I was looking for a project that brings these two fields together and the TRIFECTA project turned out to be exactly the right project for that. My thesis, which is supervised by dr. Marieke van Erp, focuses on relation extraction for Dutch maritime history. I applied natural language processing (NLP) methods to automatically extract and classify relations between entities such as ships, persons and commodities from historical maritime texts.

Research on digital methods for Dutch maritime history is quite limited, which means that there are a lot of opportunities to contribute something new. For my thesis, my goal was not only to focus on the NLP side, but also to make the work useful for historians, which is very much in line with the TRIFECTA project. I modified an existing dataset and annotated it with relations (on which language models would later be trained) to address questions from historians. Digital humanities projects are inherently collaborative and require knowledge from various fields, such as history, linguistics and computer science. Luckily, at the KNAW Humanities Cluster, there are a lot of experts from various backgrounds with whom I could talk to.

Photo of the historical 17-20th century buildings housing the KNAW HuC office in downtown AmsterdamThis collaborative environment is also reflected in the day-to-day at TRIFECTA. There are weekly meetings with the team. The team consists of academics from various backgrounds, some working on NLP, other on semantic web, and others more on the history side. I’ve found these meetings very informative. The meetings gave me insights into what colleagues are working on, into the academic environment, and the opportunity to ask questions. There are also monthly research meetings with the DHLab, where colleagues present their work. As a student, you can (and are encouraged) present there. I will be rehearsing my thesis defense at the next research meeting, which is very useful since you’re already presenting to academics. I also have weekly meetings with my supervisor, Marieke, where I discuss progress and ask questions related to what I’m working on. Besides the meetings, the DHLab organizes a monthly NLP reading club where we read about topics in NLP and have hands-on sessions where we put into practice what we have read. Participants bring their own dataset or work to these sessions. The team also attends relevant digital humanities events together. The last one was a Cultural AI event where master’s students presented their thesis projects on AI applied to cultural heritage. Through all these meetings, I got familiarized with the research community and how it operates.

Everyone at TRIFECTA, and the DHLab and KNAW Humanities Cluster is super helpful and approachable, creating an inspiring and supportive environment. I would highly recommend any student to write their thesis within an interdisciplinary research project.

Categories
Research Visit

A research visit at King’s College London in September

In September, TRIFECTA PhD Candidate Jiaqi visited two researchers Dr. Albert Meroño-Peñuela and Dr. Barbara McGillivray at King’s College in London for two weeks.

Jiaqi would like to explore semantic change detection for her thesis and this paper co-authored by King’s researchers looks very interesting to her (https://aclanthology.org/2022.evonlp-1.7/). Thanks to her supervisor Marieke’s support, she spent two weeks sitting in the postgraduate office of King’s Department of Informatics with other PhD students there, working together with Dr. Nitisha Jain, and exploring London after work (e.g., a West End musical, a jazz club live in Soho, a classic movie at Prince Charles Cinema, and museum visits etc).

After introducing herself to Kings’ people and having discussions with Albert, Barbara, and Nitisha on how to make the best of this visit, Jiaqi started to work on improving a previous semantic drift paper finished by Albert, Barbara, Nitisha, and Jongmo by conducting topic modelling and adding annotations. Since this paper deals with semantic change in the disability domain (how disability-related words evolving societal attitudes can influence the perpetuation of harmful language that reinforces stereotypes and discrimination), Jiaqi also designed an annotation framework for the offensiveness of disability-related words. She also tried to apply Jongmo’s code to her VOC data to see what would happen.

Although it is a short visit, we are happy to proceed with what we have done during these two weeks and collaborate further in the future. It was also a nice experience for Jiaqi to experience a different work culture from the HuC and talk to other PhD peers to know what they are working on and practical things like how the PhD is structured at King’s/how they feel so far.

Fueled up again!