Identifying Representative Textual Sources in Blog Networks

Karen Wade, Derek Greene, Conrad Lee, Daniel Archambault, Pádraig Cunningham

Abstract

We apply methods from social network analysis and visualization to facilitate a study of the Irish blogosphere from a cultural studies perspective. We focus on solving the practical issues that arise when the goal is to perform textual analysis of the corpus produced by a network of bloggers. Previous studies into blogging networks have noted difficulties arising when trying to identify the extent and boundaries of these networks. As a response to calls for increasingly data-led approaches in media and cultural studies, we discuss a variety of social network analysis methods that can be used to identify which blogs can be seen as members of a posited “Irish blogging network” (and hence sources of textual material). We identify hub blogs, communities of sites corresponding to different topics, and representative bloggers within these communities. Based on this study, we propose a set of guidelines for researchers who wish to map out blogging networks.

Introduction: The Sampling Problem in Digital Humanities

Researchers in the humanities and cultural studies often struggle with the sheer scale of social media data. When attempting to perform deep qualitative or ethnographic analysis, they frequently fall into two traps: arbitrary sampling based on personal familiarity, or popularity bias, which focuses exclusively on "A-list" bloggers with the most followers. These methods often result in impoverished analyses that fail to capture the true boundaries and internal diversity of a network. This paper presents a data-driven framework for identifying a representative sample of blogs within a specific national "blogosphere" (Ireland), moving beyond simple popularity to uncover niche communities and "local heroes."

Methodology: Mapping the Irish Blogosphere

The researchers employed a multi-stage pipeline to map the network. They began with a snowball sampling of 21 "seed" blogs, crawling their blogrolls two steps out. After manual filtering for activity and geographic relevance, they settled on a core set of 614 blogs. This Irish Blog Corpus contains approximately 176,000 unique posts spanning from 1997 to 2011.

The network was analyzed through two distinct lenses: static "blogroll" links representing long-term affiliation, and dynamic "post-links" found within actual blog content. To group these blogs by content, the researchers used Ensemble Non-negative Matrix Factorization (NMF) on a TF-IDF term-profile matrix. This "soft" clustering approach allowed blogs to belong to multiple topical communities simultaneously, reflecting the multifaceted nature of personal blogging.

Identifying Topical Communities

The NMF algorithm successfully identified 12 distinct thematic clusters within the Irish blogosphere. These ranged from broad categories like "Discussion" (covering politics and culture) to highly specific niches such as "Gaelic," "Wine," and "Fashion."

Topical clusters identified in the Irish blogosphere using ensemble NMF.

By applying social network analysis within these clusters, the researchers identified "local hubs"—nodes with high in-degree centrality within their specific community. This approach ensures that the final sample includes influential voices from every sub-culture, rather than just the globally popular sites that dominate the entire network.

Results and Validation

The study found that relying solely on global popularity provides poor coverage of network diversity. While the top 20 most-linked blogs shared significant overlap between link types, they ignored entire sub-communities. Furthermore, blogroll links proved to be higher-quality indicators of community than dynamic post-links, as blogs connected via blogrolls showed higher vocabulary similarity.

To validate these findings, the researchers compared the text-based clusters against external tagging data from the social bookmarking site Del.icio.us. The results showed strong consistency, particularly in niche areas.

Validation of text-based clusters using external Del.icio.us tag data.

The final output of the study was a list of 14 representative blogs that provided a balanced map of the Irish blogosphere. This list was gender-balanced and covered diverse authorial stances, ranging from professional journalism to personal diaries.

Conclusion: Beyond the "A-List"

This research demonstrates that a truly representative sample of a digital network should consist of the hubs of individual communities rather than the hubs of the entire network. While challenges remain—such as the manual effort required for filtering and the difficulty of capturing isolated "personal" blogs—this framework provides a robust, reproducible method for humanities researchers to navigate the complexities of large-scale social media data. Future work will focus on temporal dynamics and the development of integrated tools to make these advanced analytical methods more accessible to non-technical researchers.

Summary updated

Abstract

We apply methods from social network analysis and visualization to facilitate a study of the Irish blogosphere from a cultural studies perspective.We focus on solving the practical issues that arise when the goal is to perform textual analysis of the corpus produced by a network of bloggers. Previous studies into blogging networks have noted difficulties arising when trying to identify the extent and boundaries of these networks. As a response to calls for increasingly data-led approaches in media and cultural studies, we discuss a variety of social network analysis methods that can be used to identify which blogs can be seen as members of a posited “Irish blogging network” (and hence sources of textual material). We identify hub blogs, communities of sites corresponding to different topics, and representative bloggers within these communities. Based on this study, we propose a set of guidelines for researchers who wish to map out blogging networks.
BibTeX
@inproceedings{wade2021identifying,
  author = {Karen Wade and Derek Greene and Conrad Lee and Daniel Archambault and Pádraig Cunningham},
  title = {{Identifying Representative Textual Sources in Blog Networks}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2021},
  doi = {10.1609/icwsm.v5i1.14096},
}
Figure