Source-linked AI summary
Tagging, Folksonomy & Co - Renaissance of Manual Indexing?
Jakob Voss
TL;DR
The paper addresses how tagging should be understood amid the growth of user-generated Web annotation and its contrast with traditional knowledge organization systems. It surveys tagging as manual indexing, analyzes feedback and vocabulary structures, and proposes a revised typology. It concludes that feedback blurs controlled and free indexing and that tagging systems require richer dimensions and user roles than the classic tripartite model.
Problem
Tagging is often treated as wholly new or opposed to traditional knowledge organization, leaving its relationship to manual indexing insufficiently integrated.
Method
The paper surveys tagging research, analyzes indexing and feedback processes, and presents a revised typology of tagging systems.
Results
Feedback mechanisms blur the boundary between controlled and free indexing, while tagging systems vary across knowledge-organization properties and at least four user roles.
Takeaways & Limitations
Tagging is better understood as a popular form of manual indexing whose vocabulary control and semantic relationships are variable system properties.
Abstract
from arXiv · showhide
This paper gives an overview of current trends in manual indexing on the Web. Along with a general rise of user generated content there are more and more tagging systems that allow users to annotate digital resources with tags (keywords) and share their annotations with other users. Tagging is frequently seen in contrast to traditional knowledge organization systems or as something completely new. This paper shows that tagging should better be seen as a popular form of manual indexing on the Web. Difference between controlled and free indexing blurs with sufficient feedback mechanisms. A revised typology of tagging systems is presented that includes different user roles and knowledge organization systems with hierarchical relationships and vocabulary control. A detailed bibliography of current research in collaborative tagging is included.
1 Introduction
The Web’s rapid growth made fulltext indexing standard, but structured metadata and social software have renewed manual subject indexing.
- 1 Introduction: Manual subject indexing is resurging through structured metadata, specialized search aggregation, and keyword annotation in social software.Examples include RSS, OAI-PMH, RDF, OpenSearch, and systems that let users annotate resources with keywords.
2 Tagging systems on the rise
Tagging systems let end users annotate and share descriptions of Web resources, and their rapid expansion has produced varied applications and practices. These systems are open and easy to use, but their flexibility also exposes vocabulary problems while enabling unconventional solutions and renewed research on manual indexing.
- 2 Tagging systems on the rise: Tagging shifts subject indexing from experts alone to end users whose annotations become immediately visible on the Web.The paper lists collaborative tagging, social classification, social indexing, and folksonomy as names for this practice.
- 2 Tagging systems on the rise: Tagging supports bookmarks, photos, weblogs, bibliographic metadata, library records, and hierarchical categories.Wikipedia categories function as hierarchical tags, while services such as BibSonomy and LibraryThing manage shared bibliographic metadata.
- 2 Tagging systems on the rise: Open tagging applications can expose homonym and synonym problems, while trial-and-error development also produces unconventional solutions.The paper attributes this combination to the limited theoretical maturity of some users and developers and to experimentation.
- 2.1 Research on Tagging: The popularity of tagging has generated research spanning workshops, specific applications, mathematical models, and debates about its relationship to classification systems.The cited work covers museums, enterprises, knowledge management, expert finding, and collaborative-tagging models.
3 The indexing process
Subject indexing represents a resource through conceptual analysis and translation into index terms, with vocabulary structures and feedback supporting the translation. The paper argues that feedback makes the boundary between free and controlled tagging less strict.
- 3 The indexing process: Subject indexing constructs a resource representation by deciding what it is about and translating that analysis into tags.Conceptual analysis concerns the resource and relevant concepts; translation finds index terms that represent them.
- 3 The indexing process: Indexing consistency is difficult because conceptual analysis depends on users’ needs and translation is affected by vocabulary control.Different users may focus on different aspects of the same resource, and controlled vocabularies support term selection through feedback.
- 3.1 Typology of knowledge organization systems: Knowledge organization systems range from term lists to classifications and categories, then to relationship groups with stronger semantic structure.The progression reflects increasing language control and increasingly explicit relationships among concepts.
- 3.2 Vocabulary control and feedback: Wikipedia combines collaborative tagging with a thesaurus-like category system, allowing users to reuse categories while also changing the vocabulary.Its categories are shared per article rather than independently assigned by every user, placing the system between controlled vocabulary and free keywords.
- 3.2 Vocabulary control and feedback: Feedback in collaborative tagging can move behavior toward consensus, so free keyword tagging and controlled-vocabulary tagging are not sharply separate.The paper describes the resulting emergent folksonomy as emergent semantics or wisdom of the crowds, while noting that crowds are not always wise.
4 Typology of tagging systems
The revised typology treats tagging systems as combinations of properties concerning aggregation, vocabulary, connectivity, automation, resources, rights, feedback, and user roles. It replaces the simple resource–user–tag model with a richer account of system variation and overlapping responsibilities.
- 4 Typology of tagging systems: The revised typology asks how tags aggregate, how vocabularies are controlled and connected, and how resources are linked and represented.Its questions distinguish set- from bag-model aggregation, several relation structures, resource hierarchies, and resource presentation.
- 4 Typology of tagging systems: It also examines automatic tagging, tagging rights, resource sources, and feedback mechanisms that suggest, display, or reject tags.These dimensions cover automatically generated terms and relations, restrictions on contributors, resource provenance, and interface support for tag entry.
- 4 Typology of tagging systems: The classic resource–user–tag model is too simple because tagging systems contain varied resources, tags, and user roles.The paper identifies resource authors, resource collectors, indexers or taggers, and searchers.
- 4 Typology of tagging systems: The four roles are resource author, resource collector, indexer or tagger, and searcher.A resource author creates or edits; a collector adds resources; a tagger assigns tags; and a searcher uses tags to find resources.
- 4 Typology of tagging systems: Roles often overlap across systems and can change over time for the same person.Examples include private bloggers combining authorship, collection, and tagging, and Wikipedia authors combining authorship with collection or tagging.
5 Conclusion
The paper frames collaborative tagging as a popular form of manual indexing rather than something fundamentally new. Feedback and broader system features blur the boundary between uncontrolled and controlled indexing.
- Tagging systems let participating users annotate digital objects with free keywords and immediately share their annotations.Instant visibility both motivates participation and helps establish feedback mechanisms.
- Feedback mechanisms make the drawbacks of uncontrolled indexing less dramatic and blur the border between controlled and free indexing.
- Vocabulary control and relationships between index terms can characterize both tagging systems and traditional knowledge organization systems.
- The paper argues that collaborative tagging is popular because non-experts can add keywords, but it is not fundamentally new.
- Further research should examine when features such as vocabulary control are needed and how they influence tagging behavior and system evolution.