Profile-based object matching for information integration

Traditional object-matching methods rely on similarities among shared attributes. Profile-based object matching builds on this approach but also correlates disjoint attributes to improve matching accuracy. To illustrate the PROM approach, we use two relational tables: one contains information about movies, the other about movie reviews.

Authors

• AnHai Doan
• Ying Lu
• Yoonkyong Lee
• Jiawei Han University of Illinois Urbana-Champaign

Keywords

• PROM
• Motion pictures
• Databases
• Springs
• Information management
• Warehousing
• Data mining
• Artificial intelligence
• Merging
• Government
• Object Matching
• Information Extraction
• Data Warehouse
• Matching Accuracy
• Film Criticism
• Sanity Check
• Training Data
• Similar Estimates
• Precision And Recall
• Confidence Score
• Domain Experts
• Training Tasks
• Matching Task
• Hard Constraints
• Task Domain
• Matched Filter
• Runtime Efficiency