Taxonomies: Foundational to knowledge management
A great deal of effort often goes into collecting information and producing reports, but they sometimes languish or end up in an archive without having been mined. In order to get the most value, the content should be enriched with metadata and made available for other purposes, including generative AI (GenAI). “Frequently, organizations conduct studies that generate a lot of data that goes into reports,” said Joseph Busch, founder and principal of Taxonomy Strategies, LLC. “Once the reports are finished, they can either become static reference documents or sources of additional insights. The more you can process them, the more value can be extracted. However, it’s essential to provide the taxonomy and metadata that support analytics and generative AI in order for this to happen.” Taxonomy Strategies assists organizations in developing metadata and taxonomy strategies to better organize their content to solve their business problems.
The documents resulting from these studies can be used to build collections that can in turn use retrieval augmented-generation (RAG) to enhance retrieval. This provides the ability to use the information both for GenAI and effective information retrieval. “The content in research reports is often very rich, and it is an underutilized resource in many cases,” Busch noted. “Not many organizations are good at doing this yet; they are interested but not necessarily making it a priority. We design the taxonomy and metadata to allow for this in the future, however.”
Getting started
As with any implementation, upfront planning is essential. “A lot of the challenge starts with defining a solvable problem and scoping it appropriately,” said Busch. The company provides metadata frameworks and taxonomy strategies to corporations and government agencies. “You need to know what problem you are trying to solve. Otherwise it is difficult to figure out what information and terminology you need,” he stated. The ideal individual for this role is a business analyst who understands what the organization needs. “Once the vision is achieved, then a team can be formed which may or may not include a taxonomist; a records management or subject matter expert might fill this role.”
Sometimes organizations try to cover too much initially, according to Busch. For example, a company might want a business taxonomy to organize all of its information. “If a company tries to use a general-purpose business taxonomy out of the box, it might not deliver a lot of value, especially if the intended use is generative AI,” he continued, “because in the absence of relevant information, it is likely to create hallucinations. Even for browsing and searching, the taxonomy and meta-data need to be connected to enterprise information.” The organization should start with a specific issue, work toward presenting a solution, and then expand to other use cases. Typical objectives to start with include improved customer service or worker productivity, which can be enhanced by better access to enterprise information.
Professional support from companies with expertise in taxonomy helps bridge the knowledge gap as organizations implement these systems. Enterprise Knowledge works closely with clients to engage them in the development process. “We provide workshops and pull away the jargon to show them how taxonomy translates to something they can use and explain the business value it can provide,” explained Hedden.