[Paper Review] Of course we share! Testing Assumptions about Social Tagging Systems

Stephan Doerfel, Daniel Zöller+4 more|arXiv (Cornell University)|Jan 3, 2014

Recommender Systems and Techniques30 references3 citations

TL;DR

This paper tests four core assumptions about social tagging systems—social sharing, retrieval utility, equality of users/tags/resources, and popularity alignment—using six years of anonymized server log data from the real-world system BibSonomy. The study reveals that while some assumptions hold partially, others, such as equality and retrieval, are significantly challenged by empirical evidence, highlighting the need for data-driven validation in social tagging research.

ABSTRACT

Social tagging systems have established themselves as an important part in today's web and have attracted the interest from our research community in a variety of investigations. The overall vision of our community is that simply through interactions with the system, i.e., through tagging and sharing of resources, users would contribute to building useful semantic structures as well as resource indexes using uncontrolled vocabulary not only due to the easy-to-use mechanics. Henceforth, a variety of assumptions about social tagging systems have emerged, yet testing them has been difficult due to the absence of suitable data. In this work we thoroughly investigate three available assumptions - e.g., is a tagging system really social? - by examining live log data gathered from the real-world public social tagging system BibSonomy. Our empirical results indicate that while some of these assumptions hold to a certain extent, other assumptions need to be reflected and viewed in a very critical light. Our observations have implications for the design of future search and other algorithms to better reflect the actual user behavior.

Motivation & Objective

• To empirically test widely held assumptions about social tagging systems that have previously lacked validation due to limited access to detailed usage data.
• To investigate whether social tagging systems are truly collaborative, retrieval-oriented, equitable across users/tags/resources, and consistent in popularity metrics.
• To provide a methodological framework for testing such assumptions using server log data, enabling future comparative studies.
• To contribute a unique, anonymized dataset of BibSonomy logs to support further research on user behavior in social tagging systems.
• To challenge existing theoretical models in social tagging by grounding them in actual user request and posting behavior rather than self-reported or post-only data.

Proposed method

• Collected and analyzed six years of anonymized server log data from BibSonomy, capturing both posting (tagging) and requesting (page views) actions.
• Used request logs to measure actual retrieval behavior, contrasting it with tagging behavior to evaluate the retrieval assumption.
• Quantified popularity across users, tags, and resources by analyzing page view frequencies and request patterns.
• Compared aggregate-level popularity trends with individual-level behavior to assess the validity of the popularity assumption.
• Evaluated social behavior by analyzing the frequency and context of resource sharing and user interaction through page visits.
• Applied statistical analysis to compare user, tag, and resource page visit rates to test the equality assumption.

Experimental results

Research questions

• RQ1 To what extent do users in BibSonomy actually share resources socially, as opposed to using the system for personal information management?
• RQ2 How well does the tagging behavior of users align with their actual retrieval behavior—does tagging truly support later retrieval?
• RQ3 Is there an equitable distribution of attention and usage across users, tags, and resources, as assumed in folksonomy models?
• RQ4 To what extent is popularity in posts (e.g., frequent tags) mirrored by popularity in requests (e.g., page views)?
• RQ5 How do individual-level behaviors compare to aggregate-level patterns in terms of popularity and usage?

Key findings

• While some user actions indicate social sharing, a significant portion of tagging and browsing behavior is individualistic, suggesting that social tagging systems serve both social and personal information management purposes.
• Only a small fraction of posted resources are retrieved later, indicating that the retrieval assumption—tagging for later access—is not strongly supported by actual usage patterns.
• User pages are visited far more frequently than resource or tag pages, providing strong evidence that the equality assumption in folksonomy models does not hold in BibSonomy.
• Popularity patterns in posts and requests align at the aggregate level but show weak correspondence at the individual level, indicating the popularity assumption holds only partially.
• The disparity in visit frequency between users, tags, and resources suggests that user-centric design and popularity-based ranking may be more effective than symmetric treatment of all three entities.
• The study demonstrates that algorithms relying on symmetric treatment of users, tags, and resources—such as FolkRank—may benefit from incorporating popularity weights or request-based transition probabilities.