Weighted Double Metaphone Implementation Details

Search with the Lucene query syntax, using a 'sounds like' algorithm.  The exact user-entered text is taken into account, so correct spelling will override the 'sounds like' algorithm.  Searches on the same indexed property value as the other double metaphone search.

Algorithm:

The Weighted Double Metaphone search has the following characteristics:

• This search is case in-sensitive. 
• It searches on the double metaphone property value and the property value. 
• Preference is given to the matches with the correct spelling.   

• Parsing is done with the following analyzers:

  • dm_propertyValue - Uses our custom double metaphone analyzer.  This has the following filters:

    • LowerCaseFilter - for setting to lowercase
      
    • StopFilter - to remove stop words (the, a, etc.) from the search
      
    • DoubleMetaphoneFilter - for testing double metaphone sounds
      

  • propertyValue - Uses our custom standard analyzer that has no stop words.

   

Example of use:

The following examples are based on the Automobiles coding scheme.

Example 1:

Search string: car

Lucene query: +dm_propertyValue:KR propertyValue:car

Result: 2 results

• Result 1:
  • entity code: C0001
  • entity description: Car
• Result 2:
  • entity code: C0002
  • entity description: Kar

Example 2:
Search string: kar

Lucene query: +dm_propertyValue:KR propertyValue:kar

Result: 2 results

• Result 1:
  • entity code: C0002
  • entity description: Kar
• Result 2:
  • entity code: C0001
  • entity description: Car

Associated JUnits:

Junits can be found here: https://github.com/lexevs/lexevs/blob/master/lbTest/src/test/java/org/LexGrid/LexBIG/Impl/function/query/lucene/searchAlgorithms/TestWeightedDoubleMetaphone.java