The "Newsroom autotagger" uses a neural network to suggest terms from the site taxonomy best describing the author's draft content. The author can then validate or reject the suggested terms and add more, automatically training the autotagger further.
The result of one of a series of short projects I ran for the European Commission to explore innovative online communication tools, the neural network was trained on the Department's existing "Newsroom" database.
When the PoC was evaluated, in 80% of the cases it was found to suggest terms which the author had missed but would have found useful. This is extremely useful on most sites, where:
a) the taxonomy is large and complex
b) individual authors know only the handful of terms they use regularly.
By ensuring authors use all relevant terms, their content is more discoverable.
Given that many DGs of the EU Commission use the Newsroom application, this could be a relatively easy way of propagating AI-supported knowledge management across the EC.
More Stuff I Do