synonym finder text-mining algorithm

Question

I would like to create an automatic synonym finder algorithm (mostly for brand names). For example, if the user enters the word "Coca cola", I would like to return the word "Coke".

This can easily be done using dictoniaries; but than I need a dictonary file for that. How can I automaticly create one? (I have a small corpus with 1 million+ records, the synonyms will come from this).

Are there any text-mining algorithms for this problems? Also, there are some online synonym services for this, but these contain no brand names :-(

Thanks, krisy

What do your records look like? Can you cluster your records to find similar terms? — Rob Neuhaus, Jun 21 '12 at 20:34

score 0 · Answer 1 · answered Jun 18 '12 at 10:29

0

A trie based algorithm can find synonyms or similar words. It's not using keys but position in the tree to define the key. For example from A you go to B and from B you go to C. There are many data structures for a trie. The most simple is an adjaceny list or a ternary search tree.

answered Jun 18 '12 at 10:29

Micromega

12,486
7
35
72

How do you suggest finding "Coke" for "Coca cola" will work with tries? Ties would be greate for similary words/misspelled words, but I don't see how they can be used for synonims :-( – krisy Jun 18 '12 at 10:46
Show the first matched trees under letter C? – Micromega Jun 18 '12 at 10:48
I fear there would be too much words returned :-( – krisy Jun 18 '12 at 10:51

synonym finder text-mining algorithm

1 Answers1