Why should you care?

AIPredict catalog search should boost items the user is likely to buy, not only TF-IDF matches.

See it live — copy this example

Use a .NET console or Web API project with Microsoft.ML. Run dotnet run after pasting.

public class QueryDocRow { public string Query; public string Title; public float Label; }
var rank = ml.Transforms.Text.FeaturizeText("QF", nameof(QueryDocRow.Query))
    .Append(ml.Transforms.Text.FeaturizeText("TF", nameof(QueryDocRow.Title)))
    .Append(ml.Transforms.Concatenate("Features", "QF", "TF"))
    .Append(ml.Regression.Trainers.Sdca());
var rankModel = rank.Fit(clickPairs);

What happened?

• Label = click or purchase signal.
• Re-rank Elasticsearch hits with a regression or LGBM-style trainer output.

Practice next

1. Build query–title pairs from clicks.
2. Regression on relevance label.
3. Re-rank top-50 ES results.
4. Add category match feature.
5. Blend ES score with ML score.

Remember

Retrieve then ML rank. Query+title features. Limit candidate set.

AIPredict search rerank

User searches "wireless earbuds".

Outcome: Clicked SKUs rise in top-5.