Philgoo Han writeup of Cohen, Ravikumar and Fienberg

This is a review of Cohen_2003_a_comparison_of_string_distance_metrics_for_name_matching_tasks by user:Ironfoot.

• Comparison of string distance metrics
  • Open source java toolkit for name-matching
  • Little proir knowledge, ill-structured data
• Edit-distance like functions
• Token based distance functions
• Hybrid distance functions
• Blocking methods: Not practical to match all pair
• Results
  • Matching: SoftTFIDF is generally the best
  • Clustering: Token based is good in average but bad when there are many misspellings
  • Combination of distance metrics: better result but training overhead