Synonyms, stopwords, stemming and named presets shape how a query is understood before ranking ever runs — and autopilot proposes synonym fixes from real zero-result and reformulated queries, with the evidence attached.
Reformulated queries and zero-result searches surface a candidate synonym or stopword, with the query evidence attached.
Every suggestion is accepted or dismissed by a person — nothing changes retrieval silently.
Save a synonym set, stopword list and stemming config as a named preset and apply it across collections.
One-way and multi-way synonym sets, per collection, per locale.
Suppress noise words that dilute relevance without hand-tuning ranking.
Language-aware stemming so "running shoes" also matches "run shoe".
Bundle a tuning configuration once, reuse it across every collection that needs it.
Autopilot doesn't invent synonyms — it mines them from queries that were reformulated, abandoned, or only converted after a rewrite, and attaches that evidence to the suggestion. A person still accepts or dismisses every change.
{
"root": "sneakers",
"synonyms": ["trainers", "kicks", "кроссовки"]
}
// autopilot suggestion, awaiting review:
// { "type": "synonym", "root": "hoodie", "synonyms": ["худи"],
// "evidence": { "queries": 41, "zero_result_rate": 0.68 } }Explore more
The best way to evaluate a search platform isn't reading one more comparison table.
Only then make the decision.