Entities
For example, recognize brand, product type, model, material, or target group when they are explicitly in the query.
Entity RecognitionAcademy · Advanced
Query understanding: understanding what the customer means.
A query often contains multiple signals at the same time. By recognizing product type, brand, target group, properties and intention, retrieval can work more specifically.
Attributes
A question like “black waterproof jacket men’s size xl” contains multiple filter and ranking signals that can be used separately.
Attributes →Intention
A product code requires different behavior than a broad orientation question. Intention helps determine which retrieval strategy is appropriate.
Intent DetectionCustomer language
Local terms, language, and alternative formulations remain valuable domain knowledge, even when AI is used.
Synonyms →Exactly where necessary
SKUs, model numbers and brand names should not be overwritten by a broad interpretation.
Exact matches →Evaluate
Verify that the interpretation leads to more logical result sets and better judgments.
Evaluation →Design principle
Not every query needs the same amount of interpretation.
Advanced search combines exact rules, attributes, lexical retrieval and semantic signals. The best route depends on the query and the available product data.