Text & Data Analysis in the Wild

Morning Seminar:

Ash Charlton

Title: Expect the unexpected: text analysis as an exploratory tool to examine legacies of race and slavery in the early Encyclopaedia Britannica (1768-1860)

Text analysis enables new ways of viewing historical documents as data – but how do you navigate these texts and effectively explore them using digital methodologies? Ash will discuss findings from her research into legacies of slavery in the early Encyclopaedia Britannica, sharing what you can achieve with text analysis, as well as some of the challenges of working with digital historical texts and methods.

Morning Sessions:

An introduction to text analysis: What is text analysis? When and why might we use it? We will also look at where to find textual data? How do we retrieve it efficiently and ethically? How do we structure the data in a way that can be used to perform quantitative analysis?

Text analysis tools: We will cover basic text analysis methods. From basic pre-processing (stemming/lemmatization/stop word removal) to basic analysis techniques such as word frequency, keywords in context, and word clouds

Afternoon Sessions:

Text analysis: We will begin to look at text mining/NLP, covering methods such as topic modelling, named entity recognition, and text classification. We will look at how (and why) they have been developed, alongside the problems and limitations of these methods.

BYOD session: Participants will work together on ongoing research datasets provided in advance by the summer school attendees, focusing on good practices and troubleshooting.

Morning Seminar:

Dr Andrea Kocsis

Title: "More than feeling? - Emotion Detection in Dark Hertiga Reviews

Can digital humanities rewrite concepts from non-digital heritage studies? With the help of distant reading, my talk aims to re-evaluate why some heritage sites do not evoke hot cognition in visitors. Hot cognition is a form of affect, a direct emotional way in which we can interpret heritage experiences before or without thinking them over. I argued that the exhibition's curation, the story-telling, and levels of immersion play a more critical role in the hot interpretation than the time that has passed since the atrocity. To test my hypothesis, I have analysed 6000 TripAdvisor reviews about sites commemorating temporally distant tragedies, such as the Clifford Tower in York, the  Mary Rose Museum in Portsmouth, and the Medieval Massacre exhibition at the Swedish History Museum. Methodologically, I aimed to review the landscape of computational methods in heritage affect research (sentiment analysis, emotion detection, topic modelling) and compare them to close reading. My research proved that extant data, like online reviews, can be a valid alternative to onsite observation and interviews. 

Morning Sessions:

Introduction to web scraping and HTML tags: We will explore different techniques, different website structures, and how to solve common problems encountered in web scraping.

Web scraping data from a forum: We will apply the skills we learnt to look at how the history of Scotland is discussed on a history forum. 

Afternoon Sessions:

An introduction to sentiment analysis: We will cover the core concepts, tools, and techniques used to undertake sentiment analysis.  

Advanced sentiment analysis: Using the techniques learned earlier, we will use sentiment analysis to examine trends in popular feelings about the history of Scotland using the data we scraped in the morning. 

Morning Seminar:

Dr Uğur Ozdemir

Title: Beyond Binary Divides: Affective Sorting and the Multidimensional Landscape of Polarization

This paper introduces the concept of affective sorting as a means to examine affective polarization (AP) in society, expanding beyond the conventional approach that primarily focuses on political party affiliations. Using The American National Election Studies spanning 2004 to 2020, we observe a significant realignment in the American political landscape, characterized by a deepening of partisan identities and the alignment of sentiments towards diverse social and political groups within this partisan dichotomy post-2012. This study reveals that affective sorting offers a nuanced understanding of AP, showing how political, ideological, and social identities interconnect within a singular affective dimension, reflecting a complex transformation in the electorate's affective orientations. Our findings highlight the importance of considering the multidimensional aspects of social identities and the intricate dynamics of polarization, underscoring the profound impact of partisan identity on the electorate's affective dispositions towards various groups.

Morning Sessions:

Data analysis: Regressions. We will introduce a variety of regression models that are used for understanding the relationship between variables in a dataset. Thereafter, we will deal with linear models and generalised linear models.

Data analysis: PCA. We will explore Principal Component Analysis, a method of identifying variables with the greatest variance in a dataset. 

Afternoon Sessions:

Data analysis: Clustering. The afternoon will see the introduction of k-means and hierarchical clustering as methods of finding patterns in your data. 

Data analysis: NULL hypothesis testing. We will use this technique to test the statistical significance of observations in our dataset. 

Morning Seminar:

Dr Alexis Pister

Title: Designing Network Visualisations for the Real World 

Networks are used in a wide range of domains to model real-world phenomena such as trophic networks, financial transactions, social relationships and more. Visualization is very useful and often required to show a network's general structure and reveal potential interesting patterns. 

However, designing usable and useful network visualisations is not trivial, as the design space is vast, and many pitfalls exist such as choosing the wrong layout or using an inefficient visual encoding, that can rapidly lead to unreadable “hairball” node-link diagrams. Choosing the correct visualisation technique, layout, and visual designs is especially not trivial since networks can take many different forms: multivariate, bipartite, temporal, spatial, etc. 

In this talk, I will reflect on the field of network visualisation through the presentation and discussion of several network visualisation projects I worked on in collaboration with historians, social scientists, and epidemiologists using real data. I will discuss visualisation techniques and design choices that can lead to efficient network visualisations usable in the real world to answer research questions and reveal interesting insights.  

Morning Sessions:

Data visualisation: We will look at the basic principles of data visualisation, followed by an introduction to the "Grammar of Graphics" and various plots within this package. We will look at ways to visualise both structured and unstructured data. 

Afternoon Sessions:

Data visualisation: We will shift to advanced data visualisation, using spatial data. We will work with examples of real-world data, and cover the fundamentals of using spatial data and visualising it effectively.

Next Steps: In the final session of the summer school, together with the attendees of the other stream we will discuss the results of the week and which would be the next steps you can take to continue developing your computational skills.