Using word vector models to trace conceptual change over time and space in historical newspapers, 1840–1914

in Digital Humanities Quarterly (2022), 16(2),

Linking large digitized newspaper corpora in different languages that have become available in national and state libraries opens up new possibilities for the computational analysis of patterns of ... [more ▼]

Linking large digitized newspaper corpora in different languages that have become available in national and state libraries opens up new possibilities for the computational analysis of patterns of information flow across national and linguistic boundaries. The significant contribution this article presents is to demonstrate how word vector models can be used to explore the way concepts have shifted in meaning over time, as they migrated across space, by comparing newspapers from different countries published between 1840 and 1914. We define a concept, rather pragmatically, as a key term or core idea that has been used in historical discourse: an abstraction or mental representation that has served as a building block for thoughts and beliefs. We use historical newspapers in English, Finnish, German and Swedish from collections in the UK, US, Germany, and Finland, as well as the Europeana collection. As use cases, we analyze how the different conceptual constructs of “nation” and “illness” emerged and changed between 1840 and 1920. Conceptual change over time is simulated by creating a series of overlapping word vector models, each spanning ten years. Historical vocabularies are retrieved on the basis of vector space proximity. Conceptual change across space is simulated by comparing the historical change of vocabularies in newspaper collections from different nations in several languages. This computational approach to conceptual history opens up new ways to identify patterns in public discourse over longer periods of time and across borders. [less ▲]

Machine Learning to Geographically Enrich Understudied Sources: A Conceptual Approach

in Rocha, Ana; Steels, Luc; van den Herik, Jaap (Eds.) Proceedings of the 12th International Conference on Agents and Artificial Intelligence - Volume 1: ARTIDIGH (2020)

This paper discusses the added value of applying machine learning (ML) to contextually enrich digital collections. In this study, we employed ML as a method to geographically enrich historical datasets ... [more ▼]

This paper discusses the added value of applying machine learning (ML) to contextually enrich digital collections. In this study, we employed ML as a method to geographically enrich historical datasets. Specifically, we used a sequence tagging tool (Riedl and Padó 2018) which implements TensorFlow to perform NER on a corpus of historical immigrant newspapers. Afterwards, the entities were extracted and geocoded. The aim was to prepare large quantities of unstructured data for a conceptual historical analysis of geographical references. The intention was to develop a method that would assist researchers working in spatial humanities, a recently emerged interdisciplinary field focused on geographic and conceptual space. Here we describe the ML methodology and the geocoding phase of the project, focussing on the advantages and challenges of this approach, particularly for humanities scholars. We also argue that, by choosing to use largely neglected sources such as immigrant newspapers (a lso known as ethnic newspapers), this study contributes to the debate about diversity representation and archival biases in digital practices. [less ▲]

Mining ethnicity: Discourse-driven topic modelling of immigrant discourses in the USA, 1898-1920

in Digital Scholarship in the Humanities (2019)

This article aims to offer a methodological contribution to digital humanities by exploring the value of a mixed-method approach to uncover and understand historical patterns in large quantities of ... [more ▼]

This article aims to offer a methodological contribution to digital humanities by exploring the value of a mixed-method approach to uncover and understand historical patterns in large quantities of textual data. It refines the distant reading technique of topic modelling (TM) by using the discourse-historical approach (DHA——Wodak, 2001) in order to analyse the mechanisms underlying discursive practices in historical newspapers. Specifically, we investigate public discourses produced by Italian minorities and test the methodology on a corpus of digitized Italian ethnic newspapers published in the USA between 1898 and 1920 (ChroniclItaly—Viola, 2018). This combined methodology, which we suggest to label ‘discourse-driven topic modelling’ (DDTM), enabled us to triangulate linguistic, social, and historical data and to examine how the changing experience of migration, identity construction, and assimilation was reflected over time in the accounts of the minorities themselves. The results proved DDTM to be effective in obtaining a categorization of the topics discussed in the immigrant press. The changing distribution of topics over time revealed how the Italian immigrant community negotiated their sense of connectedness with both the host country and the homeland. At the same time, without jeopardizing the analytical depth of the findings, the method proved its value of minimizing the risk of biases when identifying the topics which stemmed from the results rather than from preconceived assumptions. [less ▲]

The GeoNewsMiner (GNM), 1898 – 1920

Software (2019)