Acmepedia: A Case Study in Wikis, Facets and KM
Faceted navigation runs on a fuel of metadata, or tags, which supply the categories users select as they navigate and filter (e.g. who wrote the document? which department produced it?) In the classical KM world, that metadata comes from several places. It comes from subject tags applied by KM experts in a categorization or taxonomy process. It can be added by machines through techniques like entity extraction. And rich metadata is imparted through enterprise process applications like CRM and ERP, coming from information in fields or passively as information flows through process steps.
With wikis, there is no central editor to manage a uniform tagging process. However, the analog to group editing in wikis is group tagging, also known as folksonomy—or to dredge up the analogy section of the old SATs, wiki:content:: folksonomy:categorization. Folksonomies are the familiar tags in Flickr or del.icio.us that each user applies to her photos or blogs. Let’s call that the purist view of a "bottoms-up" folksonomy. But without central control, vocabulary drift can bring chaos.
In our own experimentation, we found that the purists’ folksonomy can be amended very successfully with a pragmatic approach. It can succeed by blending elements of tops-down and bottoms-up organization. And we found precedent for that in the Wikipedia itself, where the commonly held notion that the content is all created bottoms up isn’t quite right. In fact, more than 1,000 trained editors patrol the Wikipedia, curating the group-written content with editorial standards.
How does this semi-structured tagging look in practice? You employ techniques that guide users:
- Controlled vocabulary: instead of free-form document tagging, first prompt users to select common terms from a controlled vocabulary, like names of industries, products, customers and geographies.
- Enter tags in facets: Instead of prompting for tags in a single field, offer faceted fields—again, like industries, products, etc. The name of the facet itself adds valuable structure.
- Auto-complete terms: Prevent vocabulary drift by using a type-ahead search to suggest known terms as the user types.
- Put an editor in the workflow queue: Actively prevent vocabulary drift with an expert. It takes less effort than you’d expect. You won’t catch everything, but you can add common synonyms and thesaurus terms, and promote frequently used terms to the controlled vocabulary.
- Auto-tag: Supplement user tagging with some based on rules, like tags derived from an author’s department or LDAP profile.
The net result of these efforts is that with a little bit of guidance, we can get far more structure from folksonomy.
Authority and Trust
Authority matters in the enterprise in a way it doesn’t on the public Internet. It’s not enough to trust that the Wikipedia is probably right when you’re certifying a regulatory filing, making a contribution to your 401k, or preparing numbers for a quarterly earnings call.
But it’s also a mistake to think that authority is binary—present or absent—and that we should discard the abundance of content not produced by a formal editorial process. In fact, trust in a source depends on the task at hand. For example, let’s say I receive a request from a reporter to interview a customer from the financial services industry, and her deadline is today. This is a complicated query, and I might need to comb through wiki content, slide presentations in the CMS, sales notes from CRM and contracts and statements of work from the ERP system to piece together an answer. I will weigh information about who is a customer and for how long, who has spoken in the past and where they are in a current sales cycle. What’s key is that people are excellent at judging matters of trust, so the trick is to provide them with as much evidence as possible to make that decision. A special type of facet, called an "authority facet," can accomplish this.
An authority facet provides precise and specific evidence about why we should trust an author or some content. Just some examples of authority facets include certification, skills, rating, years of employment, title and department. Now it’s possible to re-imagine that same query about our financial services customer, but now evaluating the content in the context of its authority facets:
- I trust field notes from the professional services department more than the sales department because they have worked more closely with the customer;
- I want to double-source a presentation written by someone in marketing that specializes in a different industry; and
- I don’t trust content written by a new employee, but I’m glad the content is there because I’ll track down his manager.
While a few tasks will still require more concrete authority, for most, this approach works. This means we can salvage less-formal content if we can help users evaluate whether it’s reasonable to use.
The Verdict
A Wikipedia for the business—an Acmepedia—is an invaluable addition to KM, provided it takes into account the material differences between the public Internet and the enterprise. But it’s hard work, and as with all KM initiatives, it requires changes in both technology and business processes. Ultimately, the computer science and information science are the easy parts—it’s the political science that’s hard.
Endeca, headquartered in Cambridge, MA, is a next-generation information access company uniting the ease of search with the analytical power of business intelligence. The Endeca Information Access Platform combines patented intellectual property, breakthrough science and a deep focus on user experience to help people find, analyze and understand information in ways never before possible. Leading global organizations like ABN AMRO, Boeing and Cox Newspapers rely on Endeca to increase revenue, reduce costs and streamline operations through better information access.