Tech Off Post

Single Post Permalink

View Thread: "Searching 'Samson'......did you mean 'Sampson'?"
  • User profile image

    Googling for Double metaphone will probably find you a better soundex - find the matching codes (primary and secondary for double metaphone, primary for metaphone), weight them according to popularity (if you have the info) and then order according to edit distance.

    Don't forget to stem the words (look for an alternative to the porter stemmer if you have time), and try and add more weight to proper nouns and nouns, a bit of weight to verbs and a negative discriminating weight to determiners.

    Then add a neural net so that you can track the best matches (based on which links the user clicked) and feedback to improve accuracy in future searches.