Algorithms

Sieve uses a multi-stage matching pipeline designed for both accuracy and performance. This section documents each stage in detail.

• Matching Pipeline
  • Pipeline Overview
  • Stage Details
  • Dual Engine Architecture
  • Performance Characteristics
• Name Normalization
  • What Normalization Does
  • Implementation
  • Pre-normalization Cache
  • Thread Safety
• Jaro-Winkler Similarity
  • Overview
  • Jaro Similarity
  • Winkler Adjustment
  • Threshold-Aware Early Exit
  • Memory Optimization
  • Complexity
  • References
• N-gram Inverted Index
  • How It Works
  • Trigram Decomposition
  • Inverted Index Structure
  • Filtering Criteria
  • Effectiveness
  • Thread Safety
  • Why Trigrams?