TL;DR
The WebStand Project introduces a customizable XML-based data warehouse designed to help sociologists analyze social networks and influence within web-based communities. By bridging the gap between raw web data (like mailing lists) and sociological analysis, the platform reveals how corporate giants like IBM and Microsoft dominate the technical standardization processes of the W3C.
What Problem Does This Paper Address?
Sociologists have long studied social behavior through interviews and reports, but the explosion of the web offers a new, electronic frontier of data: mailing lists, forums, and institutional homepages. However, this data is semi-structured and "tree-like" (think of nested email threads), making it difficult to query using traditional relational databases without complex operations like transitive closures.
Furthermore, there is a significant technical gap. Sociologists are often not experts in database languages like XQuery, yet they need to perform sophisticated temporal and relational analyses to understand how actors, their positions, and their influences evolve over time. WebStand aims to provide a platform that handles the "Extract, Transform, and Load" (ETL) cycle specifically for sociological use cases.
Methods
The authors built a platform based on Java and the MonetDB-XQuery database. The core innovation lies in three areas:
- Native Tree-Based Querying: By using XML and XQuery, the system can easily identify discussion patterns. For example, it can define a "discussion" between two people if they reply to each other multiple times within a specific thread—a task that is computationally cumbersome in standard SQL.
- User-Friendly GUI: To accommodate non-experts, the team developed a Query-By-Example (QBE) interface. Researchers can define sociological concepts, which are then translated into an XML Schema. They can then build complex queries by simply dragging and dropping elements within a visual interface.
- Source-Based Temporality: The paper proposes a novel temporal model that goes beyond simple timestamps. It tracks "provenance"—not just when an event happened, but when a specific source reported that it happened. This is crucial for sociological studies where the timing of information discovery is as important as the event itself.
Datasets
The researchers evaluated their system using a corpus of 20,697 emails posted over four years (April 2002 – April 2006) across W3C mailing lists.
- Entities: Through heuristics, they identified 2,923 unique "physical" people.
- Domains: The emails originated from 2,076 different domains (ISPs and institutions).
- Benchmarks: The study also incorporated 28 technical reports from the XQuery recommendation process to map authorship against mailing list activity.
Key Insights and Ideas
- Corporate Dominance in Standards: A "neutral" sociological analysis of the data shows that corporations heavily dominate the XQuery standard-setting process. IBM and Microsoft not only lead in post volume but also in the number of authors signing off on final recommendations.
- The "Active Few": Participation is highly skewed. Out of 3,000 participants, only four posted more than 500 messages, while the vast majority (about 2,500) posted fewer than five.
- Academic Mobilization Gap: While universities are present, they struggle to mobilize a large number of active participants compared to industry. For instance, Microsoft had only 20 unique posters but generated nearly twice the email volume of IBM, which had 35 posters.
Key Results
The platform successfully generated quantitative institutional mappings and social graphs.
One of the most striking qualitative outputs is the institutional mapping of co-authors, which visualizes the tight-knit network of industry experts who drive web standards.
Limitations and Future Work
The authors acknowledge that their temporal model is still under development and was not fully detailed in this short paper. Additionally, while the GUI simplifies querying, optimizing the performance of these automatically generated, complex XQueries remains a challenge.
Future work focuses on automating the extraction of authoring information from different versions of technical reports (moving from Working Drafts to Recommendations) to reduce the need for manual data entry. The team also aims to scale the system to handle document sizes ranging from a few kilobytes to 100MB within their 2GB XML warehouse.
Summary updated