Text Mining - Belval Campus

In 2001, an extensive project was initiated to transform the former steel production site in Belval, Luxembourg, into the Cité des Sciences. This transformation sought to revitalize a region historically characterized by heavy industry and to integrate research, education, and cultural activities within a single, modern complex. By 2015, the new campus opened, signifying a notable milestone in Luxembourg’s shift from a traditionally industrial economy to a knowledge-based one.

The architectural design of the campus preserves elements of the steel industry—such as the iconic blast furnaces—to visually connect Belval’s industrial past with its current role as a center of higher education and research. In this post, I apply text mining techniques, including a Structural Topic Model (STM), to analyze how the Belval Campus has been represented in news coverage over several years. This analysis reveals shifting themes of historical significance, transformation, and broader societal implications.

Data Collection

The data for this analysis consists of 58 news articles obtained from Google News. Since the articles were published in multiple languages, they were translated into English using DeepL to establish a uniform corpus for analysis. The text files were then standardized as follows:

Title: Title_Name
DATE: dd.mm.yyyy
.
. Text
.
Title: Title_Name
DATE: dd.mm.yyyy
.
. Text
.

This standardized format ensures that all articles could be processed systematically.

Wordcloud Analysis

The purpose of creating a wordcloud was to visualize the frequency of terms in the corpus, providing an initial overview of recurring concepts. Words displayed with larger fonts appear more frequently.

The figure above presents a wordcloud generated from the combined news corpus. Notably, terms such as “university”, “student”, “building”, and “campus” appear in larger fonts, indicating their prominence and reflecting the central role of academia in the Belval transformation. Concurrently, words such as “furnace”, “steel”, and “industrial” acknowledge the site’s heritage as a former steel production hub, highlighting the juxtaposition of historical infrastructure with contemporary educational facilities. Additional references to “restaurant”, “project”, and “2022” suggest a broader scope of local development efforts, including dining amenities and preparations for major cultural events. Overall, the wordcloud underscores the importance of the university environment while also emphasizing the ongoing transition from an industrial past to a diverse, forward-looking district.

Term Frequency-Inverse Document Frequency (TF-IDF)

To identify terms that were particularly distinctive in each year, a two-step approach was adopted. First, the number of articles published between 2015 and 2020 was examined (see the first figure below), revealing variations in media attention over time. Notably, 2019 stands out with the highest number of articles, suggesting heightened public interest in Belval’s development during that year.

Second, a TF-IDF analysis was conducted to determine which words were most characteristic of each year (see the figure below). TF-IDF stands for term frequency–inverse document frequency and is a widely used text-mining method that measures how important a word is within a specific subset of documents compared to its usage in the overall corpus. Words that appear frequently in a small number of documents but rarely across all documents receive a higher score, highlighting their uniqueness.

2015: Terms such as “listener”, “canteen”, and “restopolis” suggest early-stage coverage focusing on student facilities and everyday campus life, highlighting Belval’s emerging university environment.
2016: Words like “monaco”, “burger”, “fashion”, and “shopping” hint at expanding commercial and leisure offerings, reflecting growing public interest in new retail spaces.
2017: The prominence of “evening”, “glow”, “performance”, “furnace”, and “blast” indicates the significance of nighttime cultural events and the steel industry’s legacy as a focal point for tourism and artistic performances.
2018: References to “tram”, “lab”, “challenges”, and “tudiants” (students) point to discussions surrounding transportation infrastructure, academic growth, and the obstacles faced during Belval’s continued expansion.
2019: Terms such as “beach”, “interview”, “shooting”, “soldier”, and “residences” illustrate a broader range of social and cultural themes, reflecting a surge in reporting as Belval prepared for major events such as the 2022 European Capital of Culture.
2020: Words including “committee”, “discus”, “financing”, “planning”, and “parliamentary” suggest a shift toward policy considerations, funding mechanisms, and formal decision-making processes governing Belval’s future development.

Collectively, these TF-IDF outcomes demonstrate how media attention moved from campus-focused reporting to a more diverse discussion of Belval’s evolving social, cultural, and infrastructural landscape. This shift illustrates the campus’s growing impact on the region’s identity, transitioning from an industrial brownfield to a multifaceted center of education, innovation, and cultural engagement.

Structural Topic Model (STM)

In order to identify coherent themes within the corpus and track their evolution over time, a Structural Topic Model (STM) was applied. Unlike simpler topic modeling methods (e.g., Latent Dirichlet Allocation), the STM can incorporate metadata—in this case, publication year—into the modeling process. This approach not only reveals the latent topics but also how each topic’s prevalence shifts from one year to the next.

The figure above presents the six extracted topics along with their most salient terms. Each topic was manually labeled based on these terms and the broader context of the articles:

Campus Belval: Emphasizes the university setting, featuring words such as “university”, “student”, and “campus”.
Events: Centers on cultural or entertainment happenings, with notable keywords like “beach”, “concert”, and “party”.
Gastronomie: Reflects discussions of food and dining options, including “restaurant”, “canteen”, and “restopolis”.
Kultur/Zukunft, Esch 2022: Highlights cultural projects and future-oriented planning, especially related to Esch-sur-Alzette’s role as European Capital of Culture 2022 (terms such as “culture”, “capital”, “planning”).
Stadtentwicklung: Focuses on urban development, covering infrastructure (e.g., “tram”, “city”, “park”), transportation, and city planning.
Strukturwandel: Captures narratives around industrial-to-knowledge transformation, featuring references to the “furnace”, “blast”, and “industrial” past alongside “university” and “city”.

The second figure illustrates how these topics are distributed across the years:

Several observations emerge:

Early Years (2015–2016):
Articles often concentrated on Campus Belval and Strukturwandel, reflecting public interest in the new university’s establishment and the ongoing shift away from the steel industry.
Increasing Variety (2017–2018):
Stadtentwicklung gains more attention, indicating growing discussions about infrastructure projects such as new tram lines and urban amenities. Gastronomie also appears more frequently, suggesting that an expanding dining scene and student life began to shape public discourse.
Peak in 2019:
A clear spike in Events and Kultur/Zukunft, Esch 2022 coincides with preparations for the European Capital of Culture, driving media coverage that centered on cultural programs, festivals, and large-scale community initiatives.
Recent Trends (2020):
Focus on Campus Belval remains, but Stadtentwicklung topics persist, reflecting sustained interest in policies and planning efforts that continue to shape Belval’s evolution as both a research hub and a culturally vibrant district.

Overall, the STM reveals that while Belval’s transformation was originally framed as an industrial redevelopment story, subsequent coverage expanded to highlight student life, cultural events, gastronomic offerings, and broader urban planning. This transition underscores Belval’s emergence as a multi-faceted site where historical legacy, academic innovation, and cultural aspirations converge.

Conclusion

This text mining analysis demonstrates how news coverage of the Belval Campus has evolved over time, transitioning from early discussions centered on the feasibility of repurposing a steel production site to more diverse topics reflecting cultural initiatives, gastronomic developments, and the broader evolution of the campus as a hub for research and innovation. Although structural transformation and industrial heritage were initially focal points, subsequent coverage highlights the growth of student life, the emergence of cultural events, and ongoing urban planning efforts. As such, Belval has come to embody a multifaceted model of regional renewal in Luxembourg - one that bridges historical identity with contemporary academic, cultural, and infrastructural advancement.

For readers interested in replicating or extending this research, the complete code is available on GitHub. The repository provides the necessary scripts and instructions for data collection, preprocessing, and analysis, offering a foundation for further investigations into text mining and topic modeling.