4

When we organize content it can be done in a hierarchy information model or a flat information model. The flat model is sometimes implemented by the use of metadata in the form of tags. Tags put labels on the content which makes the information easily retrievable when the user needs to. Tags can be used as a part of filtering search queries or fixed views of a content source.

The creation of tags can be done in different ways. An organization can decide to implement tags from the top down, meaning that tags only exist if the organization representative have created it. In this model users has to ask for the creation of tags, if they need the tag. The process is slow, and users may have forgotten where the content is when the tag is available to label the content. Or worse, the content may be stored locally making it impossible to retrieve for other users.

To overcome this problem organizations can let the users create tags themselves. This process is called a folksonomy and can in a more formal way be described as “a user-generated system of classifying and organizing online content into different categories by the use of metadata such as electronic tags”. They have the advantage of being simple and easy to use and users don’t have to ask for permission upon creation. But that is also part of the problem. Unattended tags can be misused unintentionally by spelling errors, using the wrong tag or by not tagging at all. Either way, there is a problem.

Tag Cloud

Image from UCLA.edu

An organization can use both systems at the same time, having one field for managed tags and a second field for folksonomy tags. Upon retrieval of content the organization can decide priorities between these different fields using several rules. But from my experience users find these different fields difficult to understand and as a consequence, difficult to use.

This leads to the question if you’re just using one tag field where users can add tags by themselves, should these tags be governed by an organization representative? All in the good purpose of get rid of errors mentioned before? It costs time to govern folksonomy tags, and it can be seen as an investment to make the information retrieval more accurate. But is it worth the cost? And is it folksonomy tags if you govern them? Does it really matter, as long as it works?

Should folksonomy tags be governed or not?

4

This is a very interesting topic. Opinions of information scientists are mixed but we're now getting better view with more empirical research done in this field. Along with some papers on SIGCHI, I have also referred to a study by Alexis Wichowski and would be discussing this in more detail below:

Empirical study on governing folksonomy tags

A paper by P.J. Morrison in 2007 entitled “Tagging and searching: Search retrieval effectiveness of folksonomies on the Web”, concludes that when folksonomies are combined with the directories with controlled vocabularies, precision and recall results were higher than in searches using the controlled vocabularies alone. This partly means that governed folksonomy tags does help in information retrieval but it will come at cost of time. Even though it goes against the fundamental notion of self-governance, I think it's worth investing time until we come up with better methods (discussed below) because it is the best option we have to track what is happening with the non–mainstream of the information environment.

On errors in folksonomy tags

Absence of context is one of the major problems in folksonomy tags. A tag like "tv" will not pull up items tagged "television". In separate papers, Noruzi (Folksonomies: Why do we need controlled vocabulary?) and Van Damme et al (FolksOntology: An integrated approach for turning folksonomies into ontologies) have proposed using folksonomy with thesauri and ontologies. I personally think that this can be a good way to reduce errors in folksonomy tags provided we can achieve it with machine learning in near future.

Some research has already been done for developing intelligent tagging interfaces. One such technology has been discussed in this paper by Jesse Vig of University of Minnesota. They have developed machine learning algorithms that infer properties of items and users based on user-contributed content including ratings, tags, and user reviews.

Overview

With colossal amounts of information available on the web, it's worth an effort to govern folksonomy tags because they offer insight into user behavior, offer a low cost alternative and engender communities. I am confident that we will develop better NLP and ML technologies in near future so that we can auto-govern folksonomy tags.

  • Better papers are available at ACM Journal of Information Science – Adit Gupta Jul 23 '15 at 13:09
  • Most welcome :) – Adit Gupta Jul 25 '15 at 11:59
1

Consider a folksonomy as an informal taxonomy (for an individual, group, community, organization, industry, etc., entity), as opposed to a governed or formal taxonomy of the entity. Folksonomies tend to be built from ad-hoc contributions of terms, not governed choices, so in general, folksonomy tags would not be governed, although they could be increasingly governed if an increasing control of the entity's vocabulary is sought. So the answer would seem to be a range of governance based on the question - "It depends on how controlled the entity's vocabulary needs to be."

Both ungoverned and governed taxonomies could be presented to the users as alphabetical lists, categorization structures (Broader Than/Narrower Than - BT/NT, network structures (Related To - RT), and relevance structures (e.g., sorts, clouds).

The words and thus terms for a taxonomy could be extracted from content, or could be individually entered or imported.

I have seen the following types of progression in term governance: 1. Users enter content, and thus content words, keywords, terms, without constraint, as content authors. 2. Users enter keywords in an open alphabetical list, and provide term definition(s) and definition sources and authorities. 3. Users enter terms in an open and ungoverned taxonomy, and provide term definition(s) and definition sources and authorities. 4. Users add or request addition (depending on permissions) terms in a closed and governed taxonomy, providing term definition(s) and definition sources and authorities. 5. Based on provided definitions, Taxonomists migrate terms from source items 1-3 into destination item 4. 6. Users enter or refine content vocabulary from terms and definitions out of the governed taxonomy from item 4.

Items 1 through 6 provide a continuous content refinement and vocabulary alignment activity.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.