Keep up with all of the essential KM news with a FREE subscription to KMWorld magazine. Find out more and subscribe today!

How to Use AI to Solve Business Problems Without Taxonomies (Video)

The advent of unsupervised machine-learning algorithms makes it possible for content owners to index their content without a taxonomy. This means that publishers are faced with this challenge, according to Michael Upshall, UNSILO head of business development: Maintain existing taxonomies or replace them with a full ML approach? Or, is there any way of combining the two?

In his presentation at Text Analytics Forum 2018, co-located with KMWorld, Upshall explored machine learning as an alternative to taxonomy-based approaches. Upshall looked at some case studies that have implemented different solutions, including publishers with private taxonomies used by organizations, and the use of large-scale, public-controlled vocabularies such as MeSH.

If you are looking to solve business problems using AI tools, Upshall says, would always recommend that you start with a business use case. “Start with the goal of what you’re trying to do rather than focusing on the particular technology you’re using to fix it.”

Then, he said, “Having decided the use case, then choose the best tool for the job. Understand--and this is a role for everyone in this room—it’s not the role of the technologists who build this stuff. It’s the role of you guys to decide what’s the most appropriate technology, what the machine can do well, what the humans can do well, and then identify a solution which combines a human and machine tools in an effective way.”

It is always a good idea to have some human intervention, he added. “You notice that we’re using an entirely automated process, then adding a human layer for individual solutions which makes the humans comfortable. They can play a part and they can be involved in the process. Finally, equally importantly, advice on how to use and evaluate. The standard evaluation measures for this kind of thing may not be what you are familiar with, as in checking against a human score.”

Upshall finished with a quote from the STM Association, a leading trade association for all scientific technical publishers: There are cases where the use of a taxonomy or ontology is still appropriate, but this should no longer be the assumed starting point.

“In other words,” he said, “my recommendation is, you’re the people to identify what starting point should be. And I think that's where the exciting challenge in this area is. There’s more than one way to skin a cat, and it’s for you to decide the best way to solve it.”

View the video.

Many speakers have made their slide decks available at www.text-analytics-forum.com/2018/Presentations.aspx

Learn more at Text Analytics Forum 2019, coming to Washington, DC, Nov. 6-7.

KMWorld Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues