Cultural and technological issues
Understanding the users’ mental model is critical. “The art is in divining what the users call ‘things,’ ” Hlava pointed out. “How do they think about the information?” In addition, cultural issues often surface. “The technical challenges are solvable,” she continued, “but overcoming cultural obstacles can be a challenge.” In others, branding can be a problem. “We worked with a professional services company in which we identified comparable content as across departments, but the groups had staked their reputations on term preferences they used for their content and wanted it labeled as such because it formed their community of practice.”
Finally, political or cultural viewpoints can interfere with indexing certain terms that are not considered acceptable by some groups, even if they are found in the content, especially the historical content. “Not allowing those words to be used even as synonyms hides the previous research and discussions and prevents full recall and disclosure on subjects,” Hlava noted.
Some technical challenges do remain. In order to make full use of taxonomy software, it must be integrated with other enterprise repositories and applications. “Enterprise knowledge graphs, semantics, and, most recently, LLMs and generative AI are driving renewed interest in taxonomy,” said Heather Hedden, senior consultant for taxonomy at Enterprise Knowledge. “Related to this is the realization of the need for connecting data and content across the organization and the role that taxonomies play beyond the siloed controlled vocabularies that have existed in separate applications. Enterprise Knowledge is a professional services company focused on delivering end-to-end knowledge, information, and data management services.
Different applications, including taxonomy management, tagging, content management, and intranets, need to be connected. “While the technology of APIs has existed for a long time,” Hedden added, “building connections among applications takes valuable developer time, so there is competition for resources. Culturally, semantic integration needs to be elevated from nice-to-have to a necessity.”
A semantic layer is a standardized framework for data that serves as a connector between data and knowledge. It explains the intended meaning of natural language. “The semantic layer connects all organizational knowledge assets, including content items, files, video, and other media, via a well-defined structure,” explained Hedden. Without a semantic layer, organizational data cannot be used to its maximum benefit because queries will not be understood in a way that allows a specific and appropriate answer. The semantic layer describes, manages, and shares data and applies a schema that represents relationships and hierarchies.
AI emerging
The use of AI has been growing in taxonomy, as it has in the case of many KM technologies. “Automatic tagging has been and continues to be the primary use of AI with taxonomies,” commented Hedden. “Next is the use of AI in combination with NLP [natural language processing] to extract terms from a body of content in order to enhance a taxonomy.” This technique can also be used for maintenance, to periodically check to see if the taxonomy is still valid, as content changes over time. “The use of generative AI and LLMs [large language models] is just beginning, primarily among organizations with large volumes of content to train their own LLMs,” she noted. “In all cases, however, human review is still necessary.”