Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community
Thomas Compton · 2025
Read Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community by Thomas Compton (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-01529.txt.
Introduction
Community unionism takes labor organizing beyond the workplace, into coalitions and local social action, and its forms have changed over a century. Thomas Compton compares two cases: the National Boot and Shoe Union in the 1920s and Unite Community in the 2010s and 2020s. Topic modeling, with a method called BERTopic, maps the themes in each union's discourse.
Word frequency checks join the topic analysis, and the study weighs how far each union's talk matches community unionism traits such as coalition building and grassroots engagement. Unite Community aligned more strongly with outward, social justice themes. The earlier union's corpus stressed internal administration, industrial relations, and member services, a traditional servicing model. Compton also draws method lessons, showing how modern language tools can sharpen the study of historical labor archives.
Catalog record
| CREATOR | Thomas Compton |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-01529
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community Thomas Compton Abstract This paper presents a comparative analysis of community unionism (CU) in two distinct historical and organizational contexts: the National Boot and Shoe Union (B&S) in the 1920s and Unite Community in the 2010s–2020s. Using BERTopic for thematic modeling and cTF-IDF weighting, alongside word frequency analysis, the study examines the extent to which each union’s discourse aligns with key features of CU—such as coalition-building, grassroots engagement, and action beyond the workplace. The results reveal significant differences in thematic focus and discursive coherence. While Unite Community demonstrates stronger alignment with outward-facing, social justice-oriented themes, the B&S corpus emphasizes internal administration, industrial relations, and member services—reflecting a more traditional, servicing-oriented union model. The analysis also highlights methodological insights, demonstrating how modern NLP techniques can enhance the study of historical labor archives. Ultimately, the findings suggest that while both unions engage with community-related themes, their underlying models of engagement diverge significantly, challenging assumptions about the continuity and universality of community unionism across time and sector. 1 Introduction The central purpose of this article is to explore how far the concept CU, is appropriate to two case studies. Wills and Simms (2004) and Holgate (2021) argue that unions were historically more involved with communities. Therefore, one would expect there to be evidence of CU in a historic case. Moreover, Unite Community has been identified by Holgate (2021) and Holgate et al. (2021) as a modern example of CU. These cases provide an opportunity to re-evaluate the claims made about CU and compare how far each case contains features of CU. Therefore, these two cases should provide useful information about union community engagement in two different periods. Moreover, they are two different unions. The B&S was a ‘new union’ which focused on the shoe production industry. Unite is a general union, also predominantly in the private sector. This means both cases share private sector workers. Both unions were national level. A key difference is that Unite Community is a specific part of Unite, which will be covered in more detail further on. Archives contain new possibilities with the use of sentence vector-based approaches such as BERTopic. Part of the argument of this article is that this modern approach to topic models can make historical documents easier to access for research. Just as the efforts to develop a pipeline for OCR (https://github.com/UnbrokenCocoon/OCR-evaluation), will increase the speed at which text can be extracted from images of archives. This means that the task of undertaking an evaluation of all TU monthly reports, as Hosbsbawm (1960) suggests, would be valuable, is now more viable than during his time of writing. This means that an approach that begins with the contributions of Hobsbawm (1960; 1964; 2015), that seeks to access wider amounts of textual data than he could, is more viable for future research. Figure 1 : Comparison of Top 20 Terms in Both Corpora Table 1 : Comparison of Top 20 Terms in Both Corpora Word Count bs Percentile bs Count uc Percentile uc Diff (uc–bs) branch 3058 99.96 2631 100.0 0.04 member 6584 100.0 2144 99.99 -0.01 community 155 96.74 2081 99.98 3.24 unite 37 91.05 2023 99.97 8.92 report 3013 99.96 1893 99.96 0.0 meeting 2078 99.88 1851 99.95 0.07 campaign 79 94.6 1558 99.94 5.34 leeds 222 97.62 1329 99.93 2.31 support 630 99.29 1159 99.93 0.64 union 6311 99.99 1090 99.92 -0.07 work 6032 99.99 995 99.91 -0.08 york 77 94.51 810 99.9 5.39 action 574 99.14 769 99.89 0.75 group 133 96.37 765 99.88 3.51 council 3300 99.97 733 99.87 -0.1 local 878 99.54 704 99.86 0.32 event 158 96.82 664 99.85 3.03 national 2021 99.87 658 99.84 -0.03 people 1398 99.75 620 99.83 0.08 labour 4484 99.98 613 99.82 -0.16 Table 1 : (continued) As a pre-requisite for a comparison between the two corpora, I have provided a basic table which outlines the 20 most common words shared by the corpora and sorted by Unite Community. It includes a comparison by percentile, which indicates the rank difference of each term within its corpus. Unsurprisingly, terms associated with trade union administration as shared, such as ‘branch’, ‘member’, ‘report’, ‘meeting’ and ‘national’. Each of these terms shares a similar prevalence within the corpora, suggesting that meaningful comparisons can be made because of overlapping language. These corpora are similar types of documents with significant overlap in certain terms, just as there are overlapping topics which shall be explored further on in this article. ‘Campaign’, ‘community’ and ‘event’ all score significantly higher in Unite Community. This builds on the conclusions of the latter two chapters, where, on a lexical level, it Unite Community draws more similarity with CU than B&S does. 1.1 Comparison By Percentage of Overlapping Terms in Both Corpora’s Top 20 Terms This graph provides a readable comparison between overlapping terms within the databases. It is calculated by subtracting the percentages of the top 20 most frequent terms, Unite Community – B&S. It furthers the evidence that B&S discusses ‘work’ and ‘labour’ issues more frequently than UC. This implies B&S would not meet the ‘beyond the workplace’ criteria, or at least, UC would be more likely to meet it. An issue with lists like this is that they are hard to group into useful and more thematically relevant smaller lists. This is what the BERTopic model exists to solve. This was covered in more detail in the method section. Therefore, the important arguments to make about word frequency lists are that they give insight into the presence or absence of key terms. Grouping them thematically would work at the disadvantage of the bag-of-words model. Instead, it is useful to note BS does not contain the term ‘community’ in its top 20 words, and Unite Community does. Yet, even this point is tempered by context, with chapter 5 demonstrating through bigrams that it matters what context ‘community’ is used in, especially when answering how far each case shows features of CU. 2 Topic Models Table 2 provides context of the setences which consitute both corpora. This information assists with understanding the differences in corpus size. It also demonstrates their normalisation to avoid potential cluster abnormalities caused by inproperly chunked sentences. Corpus Total Sents Avg Len Min Len Max Len < 5 Wds > 25 Wds B&S 95,557 15.22 5 270 0 4,763 UC 16,864 17.64 3 535 1 1,202 Table 2: Corpora Overview: Sentence-level statistics. Figure 2 : UMAP Projection of Unite Community Topic Model Figure 3 : UMAP Projection of B&S Topic Model 2.1 Keyword Similarity Table 3 : Thematic Groups of Topics in the Unite Community Topic Model Thematic Group Topic Labels Notes / Interpretation Trade Unions Unite Community; Trade Unions Core identity and affiliation terms. Campaign Issues Housing Issues; Climate Change; Fight the Far Right; Disability Support; Palestine Solidarity; Better Buses Campaign Outward-facing mobilisation topics reflecting CU’s social justice alignment. Administration Branch Meetings; Meetings and Working Groups; Branch Finances; Education; Local Events Internal governance and coordination; ongoing organising infrastructure. Politics Labour Party; CLP Relations Formal political relationships, party activism/motions; ties to party politics. Table 3 : Thematic Groups of Topics