Introduction to Multimodal Time Series Analysis with Python for Humanists (DHSI 2026)

Description

Data sets with rich underlying temporal dynamics are ubiquitous across the Humanities. A non-exhaustive list includes oral history interviews, medieval chronicles, news streams, diaries, as well as biographies. Even songs, poems, and novels can be approached as temporal data. Time series analysis is a branch of modern data science aiming to investigate temporal dynamics; it is an analytical and exploratory framework to study events and their relationship to time. It also offers groundbreaking solutions to explore temporal data that is unfolding through various modalities. For instance, an oral history interview is unfolding through different modalities or layers such as the textual content by the speaker, and the speaker’s body posture and eye movement. The overall goal of this course is to provide a practical introduction to single- and multimodal time series analysis tailored for humanists who have basic programming skills in Python. Specifically, our course has three goals.

First, it aims to teach the basic operations in time series analysis such as slicing, time stamping, and aggregation. As part of this, the course will guide participants through the process of transforming Humanities data sets into time series. We will discuss how existing essentially non-temporal data sets (such as a poem) can be treated as time series. We will also show how LLMs can help in the process of annotating time series and producing multimodal time series.

Second, our course will prepare participants to analyse the resulting time series. Participants will get acquainted with key concepts of time series analysis such as waiting time, recurrence, and frequency, and their connections to multimodality.

Finally, we will also teach some rudimentary statistical frameworks (survival analysis, time-to-event analysis, trend and seasonal analysis) to extract meaningful information from single- and multimodal time series.

Participants will be encouraged to bring their own data and work with that throughout the course, which will be structured as follows. In the mornings, we will offer more theory oriented sessions. By contrast, the afternoons will be devoted to practice and programming. We will provide reusable code in the format of Jupyter notebooks.

As a whole, after our course, students will be able to explore and analyse single- and multimodal temporal dynamics in a rich array of Humanities data sets, and raise meaningful questions related to the underlying temporal dynamics.

This course is for a broad Humanist audience who already has experience with Python and does data intensive research.

Instructor(s)

Gabor Mihaly Toth is a research scientist at the Center for Contemporary and Digital History at the University of Luxembourg; he is the principal investigator of “Voices from Auschwitz: Unlocking the Collective Memory with the Multimodal Analysis of Survivor Testimonies” project; before joining the University of Luxembourg, he had worked at the University of Southern California and Yale University.

Mohamed Laib is Data Science researcher in the Trustworthy AI group at the Luxembourg Institute of Science and Technology, with a solid foundation in statistics and machine learning. His work focuses on leveraging these skills to tackle complex real-world challenges.