Measuring against datasets.
We explained in the following post how our classification works:
For the taxonomy, words from articles are measured vs Classifications datasets (a.k.a. Bags of words)
To gauge how a media is perceived, we, this time, measure what is said vs Perceptions datasets
Matching and weighting are making the score of the document. This is how we know where an article belongs and its level of expertise/severity.
AI to keep datasets permanently relevant.
The perception we have on an event evolves over time. So should the classification of this article. To make it happen, we keep the datasets always up-to-date. As seen in last week post, the classification of an article evolves with time:
The evolution domino effect: Languages impact datasets, which impact taxonomy.
The way we talk about something evolves with time. In fact, the dialect of the tribe made of people concerned about something evolves over time. More specific, cooler, geekier, newer… whatever is the reason, we love changes, we love new ways to express ourselves.
Datasets updates. Permanently.
Datasets must be rebuilt and updated permanently to catch new words, new importance on words, up and down, and the disappearance of some.
Overlapping intelligence. On watch.
As datasets evolve they may make a classification overlap over another which will weaken the taxonomy.
We measure overlapping, permanently as well, and if one is detected the taxonomy evolves consequently.