[wjbrewer]

Originally Posted by starwest

Question for those of you using this mod:

Did it require much configuration to get accurate search results?

I'm having some trouble with the results returned for my search phrases. In some cases they are far less helpful, or non-existent, where the default search function found the proper products. For example, we sell products that contain Pau D'Arco Bark. On the default search, searching for "pau d arco" will return all 8 products and nothing else. With Smart Search, searching for "pau d arco" will return 50 products or more, and sorted by relevancy products with "Pau D'Arco" in their names aren't even on the first few pages.

In another instance, searching for "Chai Tea" with the default search will return 12 products which contain the phrase chai tea. Using Smart Search, of the first 10 products only 6 of them contain both words while other items like teaspoons and mate chai outrank many of the items by relevancy that contained the actual phrase "chai tea".

Hello.

This sounds like a problem with the 4 letter MySQL word limit, which would cause the incorrect results for 'tea', and 'Pau'. You can edit this by changing (or adding) this to the etc/my.cnf file:

ft_min_word_len=2

or

ft_min_word_len=3

You may have to have your web host, or web manager do this for you. After this restart MySQL and rebuild the search database and your results will be accurate down to 2 or 3 letter words.

http://dev.mysql.com/doc/refman/5.0/en/server-system-variables.html#option_mysqld_ft_min_word_len

--

Also, the Smart Search uses a natural language search. What this does is assign every word a value, and score this in relation to other words. Basically, rare words are scored higher than common words and product results are based on:

1. Word frequency.
2. Word placement (a word in the title is worth more than a word in the description).

So if you search for 'Chai Tea', the first results with be:

1. Products with both words in the title
2. Products with a single word in the title and both words in the description.
3. Products with both words in the description.
4. Products with a single word in the description.

And all based on word value and density. So if the word tea is used 10 times in a description this will rank above one that uses the work 8 times.

I hope all of this makes sense.