JevMade field notes / Worked recipe

Find a help article without forcing a false match

A search recipe compares keyword matching, Jev choosing from search results, and Jev choosing from a whole small help library.

Original by Jeroen Erne / NexibeoRetrieval

Bookmark Keep this in your collection.

Leave a noteWhat would you try with this?

Only you can see your notes.

Listen to this guide

JevMade’s plain-English explanation

0:00 /

AI narration

Independent ElevenLabs affiliate. JevMade may earn a commission.

Credits and license

“10 · Search re-ranking” 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

People often ask for help using different words from the article that answers them. This recipe compares a simple keyword search with two Jev approaches. One gives Jev a shortlist of search results; the other presents the entire small collection.

Jev receives the question and the available articles as possible answers, plus an option for none of them. The author tests all approaches on ordinary support questions, including some the collection cannot answer. This separates finding the closest article from finding one that actually helps.

A second-stage chooser cannot recover an article the first search left out. Reading everything may suit a small collection, but larger libraries face input and option limits. The reported comparison uses a small handmade dataset, so another help centre needs its own evaluation.

Key takeaways

Allow the answer that no article fits.
Check whether the initial search includes the right article.
Test realistic questions before choosing how wide a shortlist should be.

More guides →See the experiments →