Talk:Text nailing

Untitled
The section, /* TN as progressive cyber-human intelligence */ seems like advertising. Please keep this page informative, not promotional. — Preceding unsigned comment added by 142.151.170.94 (talk) 18:44, 9 October 2018 (UTC)

Request from Kartoun
Can you help embedding an image please?

http://deliveryimages.acm.org/10.1145/3140000/3139488/ins03.gif

Images included. Gknor (talk) 08:38, 5 December 2017 (UTC)

It's supervised learning...
TN is not distinct from supervised classification. It is supervised classification.

The human in-the-loop stuff is a way to generate features (the text patterns to look for). This is fair and fine, but involving domain experts to create features is not novel. Arguably it's page 1 of the book. The specific interfaces and method of bootstrapping which patterns to look for (here the initial searchwords, "toba", "cig", etc.) can be novel.

Given, the features the classifier is just a manually set threshold. Given the same features/examples almost any classifier should be able to get at least as good performance.