Topic
  • 1 reply
  • Latest Post - ‏2016-09-06T23:02:07Z by spiggle
ayanbiswas
ayanbiswas
18 Posts

Pinned topic Incorrect scores for QXNM function

‏2016-03-21T13:32:16Z | mdm-migration

I have used QXNM comparison function and ran mpxcomp job with algorithm logging enabled.

After viewing the logs ,I found out that 2 completely different tokens were getting a high score .

For instance

USEWORD[0,1]: GAICHAN              -- KISHAN               -- (mcc=P) --  +1.27
   USEWORD[1,2]: SOREN                -- SHARMA               -- (mcc=P) --  +1.27

The detailed log is attached.

My question is how can tokens like (GAICHAN , KISHAN) and (SOREN ,SHARMA) yield high scores?

Attachments

  • spiggle
    spiggle
    7 Posts

    Re: Incorrect scores for QXNM function

    ‏2016-09-06T23:02:07Z  

    You can use the madcode utility to get phonetic values for given input values.

    Both Gaichan and Kishan have a NORMPHON of KSN

    Both Soren and Sharma have a NORMPHON of SRN

     

    Input='GAICHAN', Soundex2='G250', Nysiis1='GACAN', MetaPhone='KXN', NormPhone='KSN', Prefix='GAI', ArabPhone='KSN', ArabicPhone='KSN'.

    Input='KISHAN', Soundex2='C250', Nysiis1='CASAN', MetaPhone='KXN', NormPhone='KSN', Prefix='KIS', ArabPhone='KSHN', ArabicPhone='KSN'.

    Input='SOREN', Soundex2='S650', Nysiis1='SARAN', MetaPhone='SRN', NormPhone='SRN', Prefix='SOR', ArabPhone='SRN', ArabicPhone='ZRN'.

    Input='SHARMA', Soundex2='S650', Nysiis1='SARN', MetaPhone='XRM', NormPhone='SRN', Prefix='SHA', ArabPhone='SHRN', ArabicPhone='SRM'.

     

    Whether you use phonetics in your algorithm is configurable, as well as penalties to apply in case of a phonetic match. You can also choose between a number of phonetic functions to use, but NORMPHON is the default.