Explainable results
When the AI finds a film it can't just hand back a result — it has to say why this one. Every result card carries a one-line explanation, so an editor can tell at a glance whether to trust it, and can explain to a colleague why it's being recommended.
The line on every result card
matches [music][american] · semantic+guess+literal
Read it in two halves:
First half — which of your conditions it hit: [music][american] are the dimensions from your query this film matched at once (e.g. genre, region). Dimensions it didn't match aren't tacked on.
Second half — how it was found: three sources, which can co-occur:
| Mark | Meaning |
|---|---|
| semantic | pulled in by closeness in meaning (even if the wording differs) |
| guess | the AI imagined "what the ideal film looks like," and this one matched that guess |
| literal | your query words appear directly in the title / plot |
Why this helps editors
- Decide whether to trust it: "only via guess, matched no conditions" is a signal to stay skeptical.
- Explain it out loud: a concrete basis for "why this film," not "the system said so."
- Catch problems: when results look off, this line points straight at which route misfired (see Debugging log).
More: how tags are kept from being invented and how editors review them → Auto-tagging; how it stays honest when nothing matches → Honest match.