AI Filtering (Query Tagging and Intent Classification)

What is AI Filtering?

AI Filtering automatically analyzes incoming search queries to classify shopper intent and extract named entities such as color, size, material, price range, brand, and product category. These extracted values are then applied as facet filters on the results page without the shopper having to interact with the filter panel manually.

For example, when a shopper searches “red running shoes under $100 size 10”, AI Filtering parses the query, identifies the color (red), category (running shoes), price ceiling ($100), and size (10), and passes those values directly to the search engine as active filters. The results page shows only products that match all extracted attributes from the start.

How It Works

1. The query is sent to the AI Filtering classifier at search time.
2. The classifier uses a trained NLP model to tag entity types and extract their values from the raw query text.
3. Extracted entities are mapped to your configured facet fields (color, size, price, brand, and so on).
4. The filtered search is executed and results are returned with the matching facets pre-applied and highlighted in the filter panel.

How to Configure AI Filtering

1. Log in to your ExpertRec control panel.
2. Navigate to AI > AI Filtering.
3. Toggle AI Filtering on.
4. Review the list of supported entity types and map each one to the corresponding facet field in your catalog.
5. Save your settings. AI Filtering is applied to all new queries immediately.

Supported Entity Types

• Color: Extracts named colors and maps them to a color attribute filter (e.g. “blue”, “navy”, “charcoal”).
• Size: Detects clothing sizes, shoe sizes, dimensions, and volume descriptors.
• Price range: Parses price constraints (“under $50”, “between $100 and $200”, “cheapest”) and applies min/max price filters.
• Brand: Identifies brand names present in the query and filters results to that brand.
• Material: Detects material descriptors such as “leather”, “cotton”, “stainless steel”.
• Category or product type: Classifies the broad product category the shopper intends, even when phrased informally (e.g. “kicks” to footwear).
• Intent tags: Labels the query intent as navigational, transactional, or informational so downstream logic can adjust ranking accordingly.

Examples

• womens wool coat medium grey: filters by gender (women), material (wool), category (coat), size (medium), color (grey).
• gaming laptop under 1000: filters by category (laptop), use case tag (gaming), price max ($1000).
• birthday gift for dad: classified as a discovery intent query, no hard filters applied but results are ranked toward gift-suitable products.

Best Practices

• Map all facet fields: AI Filtering can only apply filters for facets that exist in your catalog index. Make sure color, size, price, brand, and material fields are all indexed and mapped in the AI Filtering configuration screen.
• Test with real queries: Use the preview tool in AI > AI Filtering to enter sample queries and confirm the correct entities are being extracted and the right filters are being applied.
• Use alongside standard filters: AI Filtering pre-applies filters automatically, but shoppers can still adjust or remove them from the filter panel. The two mechanisms work together without conflict.
• Monitor intent classification accuracy: Review the search analytics for queries where AI Filtering was triggered. If filters are being over-applied and narrowing results too aggressively, raise the confidence threshold in the configuration to reduce false positives.