AJ
fa142288b6
Implemented word-based relevance scoring to intelligently sort transactions when multiple matches exist within the amount tolerance. The system now: 1. Splits receipt merchant name into words (handling spaces, dashes, underscores, dots) 2. Compares each word against transaction merchant name and transaction name 3. Scores based on matching word count (bidirectional substring matching) 4. Exact matches get highest priority (score 1000) 5. Word matches get scored (10 points per matching word) 6. Sorts by relevance score, then by date Examples: - Receipt "Duke Energy" matches "DUKE ENERGY CORPORATION" better than "WALMART" - Receipt "McDonald's" matches "MCDONALD'S #12345" better than "BURGER KING" - Receipt "Comcast" matches "COMCAST CABLE" better than "VERIZON" This dramatically improves auto-mapping success rate and puts the most likely transaction at the top of the manual selection list. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>