Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community

summary

The gist

This paper presents a comparative analysis of community unionism (CU) in two distinct historical and organizational contexts…

In short

The study compares community unionism in two contexts: the 1920s Boot and Shoe Union (B&S) and modern Unite Community. Using topic modeling, it found B&S focused on internal administration and industrial relations, while Unite Community emphasized outward-facing social justice themes. This shows a divergence between a traditional self-help model and a campaigning model.

Key concepts

Community Unionism (CU)
This concept examines how trade unions engage with their local communities. The paper tests whether different unions show this engagement through their language and topics, looking at things like building coalitions or acting outside the workplace.
BERTopic
This is a modern Natural Language Processing (NLP) technique used to group sentences in large text archives into coherent themes. It helps researchers find underlying topics that might be missed by simple word counting, making historical documents easier to analyze.
Thematic Alignment
This refers to how closely the language of a union matches specific community unionism features, such as grassroots engagement or coalition-building. The analysis showed Unite Community aligned better with outward campaigning themes than the B&S corpus.
Model of Engagement
This describes the fundamental way a union operates—whether it focuses on internal administration, member services, and self-help (like B&S), or on public policy campaigning and social justice (like Unite Community).

Terminology used across episodes

This episode discusses

The paper

Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community · Read on arXiv

Thomas Compton

University of York

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.

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: I'm Tom.

Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.

Tom: We're looking at the paper: "Service, Solidarity, and Self-Help".

Jane: Here's the gist of the paper. 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, revealing significant differences in thematic focus and discursive coherence regarding CU features Comparative Context Setup The study examines how far the concept of CU is appropriate to two case studies, Wills and Simms (2004) and Holgate (2021) argue that unions were historically more involved with communities The B&S was a ‘new union’ which focused on the shoe production industry, while Unite is a general union also predominantly in the private sector Both cases share private sector workers and are both national level unions A key difference is that Unite Community is a specific part of Unite, which will be covered in more detail further on Discursive Alignment via Topic Modeling The analysis uses BERTopic for thematic modeling and cTF-IDF weighting to examine the extent to which each union’s discourse aligns with key CU features such as coalition-building, grassroots engagement, and action beyond the workplace The results reveal significant differences in thematic focus and discursive coherence Unite Community demonstrates stronger alignment with outward-facing, social justice-oriented themes compared to the B&S corpus which emphasizes internal administration, industrial relations, and member services Thematic Structure Divergence Topic modeling provides a way to group semantically related sentences and extract characteristic keywords rather than relying solely on word frequency lists The B&S corpus displays greater internal cohesion and tighter clustering than the Unite Community corpus, likely stemming from the curated nature of the B&S archive For example, in B&S topics, Industrial Relations holds a more prominent position compared to UC where no distinct “strike” topic emerged The evidence suggests that Unite Community may fit into the features of Fairbrother’s (2008) criteria, with coalitions and equality having strong relevance, whereas B&S has a much weaker connection to any of the criteria The findings indicate that B&S followed a more traditional, service-oriented approach focused on ‘self-help’, while Unite Community focuses on campaigning and influencing public policy This suggests that B&S’s approach to CU differs from Unite Community’s because one focuses on methods of self-help and the other on campaigning Methodological Insights The study highlights methodological insights by demonstrating how modern NLP techniques like BERTopic can enhance the study of historical labor archives The comparison of top 20 terms shows that while shared terms exist, ‘Campaign’, ‘community’, and ‘event’ score significantly higher in Unite Community The analysis also demonstrates how the task of undertaking an evaluation of all TU monthly reports is now more viable than during Hobsbawm's time Limitations and Future Directions A limitation noted is the difficulty in proving broader community relations without access to more documents, as archives like the Labour Club lacked much appropriate information on community approaches Future research could target unions with more comprehensive archives, including letters and personal effects, to gain a broader perspective The potential advantage of the quantitative approach is that it may be more appealing to a union interested in keeping its data private Evaluation Metrics Justification The selection of the ‘all-MiniLM-L6-v2’ default sentence embedding model was justified by comparing different possible models, with ‘all-mpnet-base-v2’ being chosen The Gini Coefficient helps detect whether a model is artificially skewed, and the PUV measures how distinct each topic is The B&S model outperformed the mean across several metrics, showing superior output in terms of Gini Score and Keyword Freq Score References Hobsbawm, E.

Tom: J.

Jane: (1960).

Tom: “Records of the Trade Union Movement.” Archives: The Journal of the British Records Association, 4(23), pp.

Title and authors: Tom: So, we're looking at this paper called "Service, Solidarity, and Self-Help: A Comparative Topic Modeling Analysis of Community Unionism in the Boot and Shoe Union and Unite Community." Jane, what do you see as the big picture here?

Jane: It's comparing two very different types of unions across a long time span. We have the National Boot and Shoe Union from the 1920s against Unite Community from the 2010s to 2020s <ref:2509.01529#pg1,the National Boot and Shoe Union>. The authors are using topic modeling to see how their language actually groups up and what that says about community unionism.

Tom: Exactly. It's not just a history lesson, it’s trying to find if this concept of community unionism is even appropriate for these different eras and sectors. We're looking at the Boot and Shoe Union, which was focused on shoe production in the 1920s, and Unite Community, which is a general union mostly in the private sector today <ref:2509.01529#pg1,in the 1920s, and Unite Community>.

Lu: The paper sets up this comparison by focusing on how far unions were historically involved with communities, building on ideas from Wills and Simms and Holgate. They're trying to use data to see if the concept holds up when you look at different organizational structures.

Meng: From an engineer's view, it sounds like they are taking massive amounts of text and running it through these topic models just to find patterns in the language. It’s interesting, but I wonder if the data they used is clean enough for this kind of deep analysis.

Lalam: The authors are using BERTopic and cTF-IDF weighting alongside word frequency analysis to look at things like coalition-building and grassroots engagement. Essentially, they're trying to see if the language of these two unions aligns with those specific community unionism features.

The paper's summary: Jane: So what are the main findings from this study? Basically, the analysis shows a significant difference in thematic focus between the two groups. While Unite Community shows a stronger alignment with outward-facing, social justice-oriented themes, the Boot and Shoe Union corpus emphasizes internal administration and industrial relations.

Tom: That’s interesting because it suggests that what 'community unionism' means isn't uniform across time or sector. The B andS corpus reflects a more traditional, servicing-oriented union model focused on internal member services, whereas Unite Community is focused more on campaigning and influencing public policy.

Lu: The paper uses topic modeling to show this divergence in coherence. For example, the Boot and Shoe Union shows greater internal cohesion and tighter clustering in its topics compared to the Unite Community corpus. This difference might come from how the documents were curated or organized over time.

Meng: So if we think about deploying a model for community engagement, this suggests that a model trained on B andS data would look very different from one trained on Unite Community data. One is focused on internal operations, the other is focused externally.

Lalam: And they found that when you look at word frequencies, specific terms like "Campaign," "community," and "event" score significantly higher in the Unite Community corpus compared to the Boot and Shoe Union corpus. This gives lexical evidence for their thematic difference.

The paper's improvements: Jane: The paper suggests a few ways this research could be improved or what it opens up for future work. They point out that using sentence vector-based approaches like BERTopic is a big improvement because it makes historical documents much easier to access for researchers.

Tom: That’s true. It's about making the task of evaluating all those monthly union reports, as Hobsbawm suggested, more viable than it was back then. Modern NLP techniques are helping us access and process that kind of old data in a way that was impossible before.

Lu: They also highlight the advantage of this quantitative approach—that it might be more appealing to a union because they can keep their data private while still getting analysis done. That's a nice practical insight for future research design.

Meng: I wonder if that privacy aspect is really as scalable as they suggest, though. If you're running models on this kind of sensitive organizational text, the infrastructure costs can get high quickly when you try to scale it up for real production use.

Lalam: And they also note a limitation: proving broader community relations is hard without access to more documents. They mention that archives like the Labour Club didn't have much appropriate information on community approaches yet. Future research should target unions with more comprehensive archives, including letters and personal effects, for a broader perspective.

Conclusion: Jane: So to wrap up, the main implication is that while both unions talk about community stuff, their underlying methods of engagement are quite different. The Boot and Shoe Union followed a more traditional self-help approach focused on internal organization and member services.

Tom: And Unite Community leans toward campaigning and influencing public policy, which is what we see in those higher scores for terms like "campaign" in their discourse. It challenges the idea that community unionism has to look the same across different eras or organizational types.

Lu: The paper shows how topic modeling can be a powerful tool for this kind of comparison, helping researchers map out these thematic differences between the Boot and Shoe Union and Unite Community.

Meng: For deployment, it suggests that you need to train models specifically on the context of the union to get meaningful results. If you just use a general model, you're going to get noise that doesn't reflect what actually matters at scale.

Lalam: And they conclude that we need more data and better methods for historical work, especially when dealing with archives that don't have enough community-focused material. It’s a call for better data collection to truly understand the topic.

More episodes

← Home