Similarity Flooding

This site provides additional material for the Similarity Flooding paper.

A reference Java implementation of the algorithm (with the source code) is available for academic use. Download here. The code has been updated on Nov 17, 2003 to fix several compilation issues. (The JAR is around 300K because it also contains a portion of the RDF API). Run java com.interdataworking.mm.alg.Match and look in the source code to see how the algorithm is utilized.

New (04/28/03): A demo version of the model-management system Rondo has been released. The Match operator in Rondo is implemented using the Similarity Flooding algorithm. The system provides a GUI for examining and adjusting the results of semi-automatic schema matching operations.

The description of the nine tasks that was handed out to the users is here.

The user data collected in the study is summarized in the table below and is available for download as a zip file. The schemas used the matching tasks are available below (L/R), and are downloadable as a zip file.

  Task 1 (L/R) Task 2 (L/R) Task 3 (L/R) Task 4 (L/R) Task 5 (L/R) Task 6 (L/R) Task 7 (L/R) Task 8 (L/R) Task 9 (L/R)
User 1 1/1 1/2 1/3 1/4 1/5 1/6 1/7 1/8 1/9
User 2 2/1 2/2 2/3 2/4 2/5 2/6 2/7 2/8 2/9
User 3 3/1 3/2 3/3 3/4 3/5 3/6 3/7 3/8 3/9
User 4 4/1 4/2 4/3 4/4 4/5 4/6 4/7 4/8 4/9
User 5 5/1 5/2 5/3 5/4 5/5 5/6 5/7 5/8 5/9
User 6 6/1 6/2 6/3 6/4 6/5 6/6 6/7 6/8 6/9
User 7 7/1 7/2 7/3 7/4 7/5 7/6 7/7 7/8 7/9


Sergey Melnik
Last modified: Nov 17, 2003