JevMade field notes / Worked recipe

Give an article more than one useful tag

This recipe checks every tag separately and shows how changing the acceptance cutoff trades fewer incorrect labels for more missed topics.

Original by Jeroen Erne / NexibeoClassification

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Credits and license

“04 · Multi-label tagging” by Jeroen Erne / Nexibeo. Read the original source.

This expanded guide is an AI-narrated adaptation of the source’s essential explanation, examples and caveats, not a word-for-word reading. The synthetic voice does not imitate the author or imply their endorsement.

License: MIT

MIT License

Copyright (c) 2026 Jeroen Erne (https://nexibeo.com, https://completeaitraining.com)

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Our summary

An article can be about several subjects, so choosing one category may hide useful information. This recipe asks Jev a separate yes-or-no question for every possible tag. It checks whether each topic is a main theme rather than just a passing mention.

The program sends those questions together and receives a probability for each tag. It keeps tags above a chosen cutoff, with at least one tag retained. The author compares several cutoffs against labelled examples to show how stricter settings change the mistakes.

A higher cutoff can remove incorrect tags while also missing genuine topics. Some disagreements reflect subjective labels rather than a clear mistake. The recipe's comparison that knows the correct number of tags in advance is only an upper bound, not a usable everyday setting.

Key takeaways

Define what makes a topic central rather than incidental.
Check both wrong tags and relevant tags the program missed.
Do not mistake a comparison with advance knowledge for a practical setting.

More guides →See the experiments →