I have mixed views on tagging. On one hand, I agree that most information architecture forces the user to channel the content owner in order to find content; the inability to (easily) cross-reference data into multiple categories in most systems compounds this problem. At the same time, tagging doesn't innately solve this.
The first issue with tagging is that the arbitrary nature of it. While allowing the content owner control over tags (as is usually the case) helps with cross-referencing of information, it does little to safeguard against the categorization being heavily dependent on the content owner's perspective. Some systems mitigate this by allowing user defined tags which, on high traffic sites, solves the problem by brute force: someone out there is bound to think like you; on smaller sites, though, this only contributes to the problem by leading to false negatives or undermining usability through inconsistent tagging.
The second issue is the lack of correlation or context. While tags provide an optimal format for cross-referencing data, most tagging systems only allow the user to filter by a single tag at a time. I can filter by "telephone" or I can filter by "media player" but I can't view all items tagged with both "telephone" and "media player". In this regard, old taxonomy driven hierarchies provide more contextual organization.
Ignia recently implemented a different model on WindowsHPC.net. In this model, we implemented Artemis' Rapid Classification Framework, which allows the creation of a hierarchical taxonomy of tags. This provides the benefit of contextual associations and predefined organization of a content hierarchy as well as the ease of cross-referencing found in tags. In addition, we combined this with three techniques for exploring tags (implemented via SharePoint Web Parts) that help suit different objectives:
- Cross-Referencing: As is common practice with tags, we expose a list of tags associated with each content item, with the tags currently filtered by being removed to reduce unproductive duplication.
- Tag Hierarchy: Instead of a tag cloud, which sorts by popularity, we expose the tag taxonomy which provides not only context to tags but, more importantly, offers a familiar and intuitive approach to navigation (the treeview) - with the added benefit that the content on the backend is easily categorized into multiple "folders".
- Checkboxes: In addition to the tag hierarchy mimicking standard site navigation, it also provides an easy way to filter data based on multiple tags. The checkboxes allow a user to search for content that belongs in multiple categories.
To see this in action, take a look at the Resources page on the WindowsHPC.net site.
Jacob Nielson has advocated lengthy web pages over the traditional approach of breaking content up into multiple categories. The logic is that most users will default to the browser's find functionality. I understand (although don't fully agree with) his point, but regardless this approach provides the best of both worlds: since the tags are exclusive, the user is able to see all content initially (offering the full benefit of the browser's search feature) while making it easy to filter and cross-reference information based on tags. This suits three distinct methods of searching for data while using a single approach.
This is, by no means, the end-all-be-all to navigation or tagging. Not even close. I'll post some other time on what I feel is the ultimate endpoint of tag-based navigation. In the meanwhile, I think the approach we implemented here is a good step in the right direction and helps resolve a lot of the limitations of tags while also exposing them to users who are more comfortable or familiar with a traditional way of navigating content (e.g., via a strict hierarchy).