You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At the moment, auto-suggestions are matching words pulled from the first N matching result records. This sometimes leads to unexpected words ranking on top. E.g. (taken from the VisitPlus dataset):
"ca" -> "calshot", "castle", "camber"
castle would be the most common & intuitive suggestion. However calshot ranks higher, probably because "Calshot castle" comes up as the highest scored record in the ca search.
We could think about re-scoring suggestions based on the number of matches they would produce as a search result. This would be an extra step, and require one follow-up search per suggestion each. But if we do it, we would get castle as the highest ranked suggestions, because it produces 1k+ results as a search, while calshot only produces one result.
The text was updated successfully, but these errors were encountered:
At the moment, auto-suggestions are matching words pulled from the first N matching result records. This sometimes leads to unexpected words ranking on top. E.g. (taken from the VisitPlus dataset):
castle would be the most common & intuitive suggestion. However calshot ranks higher, probably because "Calshot castle" comes up as the highest scored record in the ca search.
We could think about re-scoring suggestions based on the number of matches they would produce as a search result. This would be an extra step, and require one follow-up search per suggestion each. But if we do it, we would get castle as the highest ranked suggestions, because it produces 1k+ results as a search, while calshot only produces one result.
The text was updated successfully, but these errors were encountered: