Folksonomies and Enterprise Knowledge Graphs
Using controlled vocabularies and an open thesaurus to democratize online content
Safe to say that the words we use help give us context.
From how we describe things on social media to how we tag products in our online ecommerce listings, words (and their related partners keywords) help form meaning and structure for online content. When culling Google for a product or page, we rely on words and click based on terms in the URL. When publishing posts, we take time to make every word right, to say what we mean, and to convey a message to potentially thousands of people.
Traditional web content is structured via a taxonomy: a predefined set of clickable terms from a dropdown menu at the top of the website that provides gateways to more information. We search a site based on context, especially the listings and terms that we’ve come to identify with that particular site or page. We often share with hashtags and other related terms to classify the content we make, find, or repurpose.
When we search Google, the knowledge panel on the right side of the screen is directly created from a culled set of web and relational data called a knowledge graph. This knowledge graph is special for Google, but it can be created and used for almost any sized business or entity, online or otherwise. Wikipedia, for instance, has its own knowledge graph-defined taxonomy with its Wikidata format and knowledge box panels on the right side of the page.
This web of structured data can be seen as an evolving template for both the Semantic Web and the proposed context graph of the internet, or a way to bring search and user intent directly into the process of browser search and online content structuring. It makes sense: a UX point of view asks users to take their own clickstreams and navigational paths and provide feedback on the usability; we might as well think of the internet and browsing more in terms of recommender engines than static pages waiting for us to make the steps happen.
A taxonomy can be formed from an ontological framework, perhaps even a controlled thesaurus with a set of terms and their semantic equivalents for use on the page. Having a controlled set of words and terms help define not only the vibe of a site or its content, but how that content relates to other content and forms navigational patterns for users.
So it can be pretty important to get this right.
Take Google Ads, for instance. The placement of key terms and phrases based on which audiences you want to reach at what stage of the buying funnel can mean the difference between success and failure; it can mean reaching only part of your audience (or none at all) or reaching precisely the right market and increasing conversions or other KPIs. Product catalogues for ecommerce can leave customers feeling ripped off if the terms used and reviews aren’t relatable or descriptive enough.
And yet, the terms we use can be jealously guarded, walled off, discovered only through rigid scientific or even cold means, such as aggregated information from an SEO program or Google Trends, keyword research or Chrome extensions that help us see what topics are trending and how people are using them. Key terms can be generated entirely in-house, which isn’t necessarily a bad thing—but there is another way to define them.
I’ve always loved the idea of community branding—not just in an outreach way, where a company establishes its brand in the community and lets users spread its tailored message, but in a cooperative way. In this way, a company’s core features are directly influenced by feedback from users, customers, or owners (in the case of a genuine co-op). Many, many companies throughout the world operate as co-ops, where the workers are owners and everyone has an equity stake in the value of the company. A brand or logo, in this case, might mean more than just the executive team and the staff, or a trademarked logo splashed on the sides of buses…it could mean something more abstract, an idea hewed to by members of a community defining a brand and its function for themselves, autonomously and then in a collective.
Similarly, for those companies operating here that want a more open brand, there is the delightful idea of a folksonomy, aka folk + taxonomy. If a taxonomy is a more rigid structure of a company and its terms built from the top down, a folksonomy is built from the ground up. Users, customers, and fans build the terms that are used to define the brand and spread a message based on an autonomous collective of sharing and remixing of identity.
For example: let’s say a Semrush keyword search has returned a set of ten target keywords for a company to focus on. These keywords can be concatenated with others for key phrases, the better to match search intent from users. These phrases and terms are used on the website to label categories, dropdowns, menus, and blog titles. The information architecture (IA) of the site is built around these researched, trend-influenced terms and its posed questions are the subjects of blog pages and a FAQ sheet.
Perhaps it’s even a little more open than this. Let’s say the company and its founders decide on the terms they want to use and whiteboard it as part of defining the brand’s voice. The voice is then a controlled vocabulary deployed on site pages as needed, with minor variations or semantic relations for navigation. As an example of a recent controlled vocab term, just look at the transition from employees to team members or teammates (or people). One is a classic but slightly colder descriptor, whereas another defines (positively) relations between working members of a team. It’s been changed in favor of a friendlier vibe.
The site pages may have been scrubbed of the term employee or employees and replaced with these friendlier terms. A good start, especially for defining a set of more positive monikers. Such changes can help positively influence and define team culture, and presents a framework for hiring and training new team members. For businesses struggling to hire and capture the voice of external reality about the company, a folksonomy and social indexing for vocabulary can help them break through, and possibly even counter possible existent criticisms.
But let’s say we opened it even more. Let’s say we didn’t just let the members of the company weigh in with terms and defined lists of vibe words. Let’s imagine we had forms, fields, and options for social tagging. Let’s say web content, web pages, blogs, and social media pieces could be tagged and defined by users and the customers themselves. These terms can then be put through a histogram or Excel file, with the most used terms ranking highest for content. These terms can be incorporated in a knowledge graph or structured data framework (such as The New York Times did), or they can be actively used for descriptions, information and knowledge classification, and lively, evolving taxonomy.
Social tagging and folksonomy used to be a more active occupation for online content. Tumblr, Digg, and others incorporated the use of social tagging of terms and categorization for user intent. Facebook, of course, still lets you tag yourself and others in pictures (if the permissions are enabled). Tagging and defining our content is a principle recipe of the internet’s ever-evolving semantic structure. Words continue to give us meaning even now.
A folksonomic classification system can also be applied to ecommerce category listings, especially letting users tag and submit their own terms for not only how they find something, but how it relates to their lives and the uses it provides. If marketing is less about the product and more about the need that product fulfills in your life, then a requested and open folksonomy process enables users to directly tag the use of these products in their lives.
Extended knowledge graphs for a community-branded business can also be built in a similar way, with many different term input mechanisms available. Questionnaires and surveys are widely used; an evolving process of social tags and keywords gleaned from sentiment analysis and shares on social networks, including Twitter and Instagram, can also help define a brand’s core terms. Part of the excitement of a lively folksonomy, however, is the way terms can always be added or changed based on directly feedback and input. A knowledge graph for an enterprise built around these terms can also include more semantic relations between them. A site or page with more interests in web accessibility can use folksonomy-gleaned ALT tags to completely democratize the process for those users it wants to reach.
Knowledge graphs aren’t only used for enterprises or Google—they have also been used for events and external documentation. COVID-19, for instance, generated a massive amount of data, papers, and scholarly work. For a team of researchers trying to use all this data, the problems can present themselves quickly. A knowledge graph attempt to reckon with all this, CORD-NER, was an open-sourced data set using named-entity recognition (NER) to build a usable, easily searchable taxonomy to enable potentially lifesaving research and insights. There were others built, but frameworks for events such as this can bring needed advances from previously jumbled and unstructured sets of data.
In online commerce sites, a product or retail taxonomy has been a boon for many. Creating a detailed, well-formatted (and knowledge graph/controlled vocabulary-influenced) catalogue for products and descriptions can help bring desperately needed organization to an unstructured but growing site.
The biggest things to keep in mind are that folksonomies are only as broad or as helpful as the users defining them. A small company might get less use out of public-facing controlled vocabulary creation than a larger enterprise or brand with multiple sources of data and indexable terms. Competition for keywords can still be fierce, and there will always be plenty of stop words and inapplicable words or too much of a broad match to be employed usefully. It comes down to a careful plucking and evaluation process, with the data cleaned and prepared before wildly reformatting a website based on social tags. Experimentation, reimagining, remixing, and open-ended design.
This is, of course, a much more complicated topic than just a simple article can do justice to. But it’s also an important opening for a new line of thinking that potentially democratizes and opens a brand, displacing it from siloed or a walled garden of controlled ideas. User-defined brand presences online can help build trust and foment a brand community that engages with, loves, and extends the work done in-house. Much like a folksonomy itself, the use of knowledge graphs and enterprise vocabularies remains an evolving, exciting prospect for a connected digital web of discovery.