Intent Handling
Intent handling detects what users are looking for in their queries (brand, category, attribute, use-case) and applies appropriate ranking behavior for each intent type.
Overview
Different query types require different ranking strategies. Intent handling automatically detects query intent and applies the most appropriate ranking behavior, ensuring users find what they're looking for.
Intent Types
Brand Intent
Users searching for a specific brand:
- Queries: "Nike shoes", "Apple iPhone", "Samsung TV"
- Strategy: Prioritize exact brand matches, show brand-specific products
- Ranking: High weight on brand match, lower weight on other attributes
{
"intent_type": "brand",
"detection_patterns": [
"brand_name + product_type",
"brand_name only"
],
"ranking_strategy": {
"brand_match_weight": 0.8,
"relevance_weight": 0.2
}
}
Category Intent
Users browsing a category:
- Queries: "running shoes", "laptops", "kitchen appliances"
- Strategy: Show diverse products from the category
- Ranking: High weight on category match, diversity across brands/attributes
{
"intent_type": "category",
"detection_patterns": [
"category_name",
"product_type"
],
"ranking_strategy": {
"category_match_weight": 0.7,
"diversity_weight": 0.3,
"apply_guardrails": true
}
}
Attribute Intent
Users searching by specific attributes:
- Queries: "red dress", "wireless headphones", "waterproof watch"
- Strategy: Prioritize products matching the attribute
- Ranking: High weight on attribute match, filter by attribute
{
"intent_type": "attribute",
"detection_patterns": [
"color + product_type",
"material + product_type",
"feature + product_type"
],
"ranking_strategy": {
"attribute_match_weight": 0.9,
"relevance_weight": 0.1
}
}
Use-Case Intent
Users searching for a use case or problem:
- Queries: "gift for mom", "workout gear", "home office setup"
- Strategy: Show products that solve the use case
- Ranking: Semantic relevance + use-case tags
{
"intent_type": "use_case",
"detection_patterns": [
"for + use_case",
"gift for",
"best for"
],
"ranking_strategy": {
"semantic_relevance_weight": 0.6,
"use_case_tags_weight": 0.4
}
}
Comparison Intent
Users comparing products:
- Queries: "iPhone vs Samsung", "best running shoes"
- Strategy: Show top products for comparison
- Ranking: High-quality, comparable products
{
"intent_type": "comparison",
"detection_patterns": [
"vs",
"compare",
"best"
],
"ranking_strategy": {
"quality_weight": 0.5,
"popularity_weight": 0.3,
"reviews_weight": 0.2
}
}
Intent Detection
Detect intent using multiple methods:
Pattern Matching
{
"detection_method": "pattern_matching",
"patterns": {
"brand": ["brand_name + product", "brand_name"],
"category": ["product_type", "category_name"],
"attribute": ["attribute + product", "color + product"]
}
}
Machine Learning Classification
{
"detection_method": "ml_classification",
"model": "intent_classifier_v1",
"features": [
"query_text",
"query_length",
"contains_brand",
"contains_category",
"contains_attribute"
]
}
Hybrid Approach
Combine pattern matching and ML:
{
"detection_method": "hybrid",
"pattern_matching_weight": 0.4,
"ml_classification_weight": 0.6
}
Intent-Specific Ranking
Apply different ranking strategies per intent:
{
"intent_strategies": {
"brand": {
"boost_exact_matches": true,
"diversity": "low",
"guardrails": false
},
"category": {
"boost_exact_matches": false,
"diversity": "high",
"guardrails": true,
"max_per_brand": 2
},
"attribute": {
"filter_by_attribute": true,
"boost_attribute_match": true,
"diversity": "medium"
}
}
}
Ambiguous Intent Handling
Handle queries with ambiguous or multiple intents:
{
"ambiguous_intent_strategy": {
"detection_threshold": 0.6,
"fallback": "category",
"show_diverse_results": true
}
}
Intent Analytics
Track intent distribution and performance:
- Intent distribution: What percentage of queries are each intent type?
- Conversion by intent: Which intents convert best?
- Intent accuracy: How accurate is intent detection?
- Intent-specific metrics: Performance metrics per intent type
Best Practices
- Start with common intents: Focus on brand, category, and attribute intents first
- Validate detection: Regularly review intent detection accuracy
- Test strategies: A/B test intent-specific ranking strategies
- Handle ambiguity: Have fallback strategies for ambiguous queries
- Monitor performance: Track how intent handling affects conversion
- Iterate: Continuously improve intent detection and ranking strategies
Related Topics
- LLM Based Ranking and Relevance - LLM-powered ranking trained on your customer data
- Personalization - User-specific ranking
- Facet Management - Help users refine intent