Behind-The-Scenes of a Board Games Website

How do you spell Carcassonne again? Is it one s and two n’s or two s’s and one n? Actually it’s two s’s and two n’s according to the Geek.

“So what?” I hear you ask - well, part of the final site builds on the current board games list to allow you to find a game. But what if you’re looking for Yinsh or Zooloretto - you don’t want to scroll through 4000+ board games before you get to the game you want to look at, so how about doing a search?

Problems with Searching

Searching through a list of games can be a little trickier when you’re looking for odd-named board games or game names in languages you’re not used to spelling. In a normal web search if you don’t type in the search text exactly as it appears in the database, then the web site’s not going to find the game - you’ll be left with a blank “No search results found” message and probably frustrated.

Web usability dictates we should help you find what you want as effectively as possible. Luckily, the underlying database and programming language I’m using to build FastMovingGames supports a little-used, but magnificent function called SOUNDEX…

A Census-tive Topic?

Digging around, it looks like SOUNDEX is an encoding system originally used to find immigrants’ names in census data when the researcher didn’t know how to spell the name (but could make an educated guess).

Mis-spelled names… getting warmer…

The SOUNDEX encoding system reduces each name to an alpha-numeric code which is one letter and three numbers. Repeat for the name you’re trying to match and you can then compare.

The system is based on how consonants are pronounced, with similar consonants grouped and encoded together. Vowels and common vowel-replacements (Y, W, and H) are disregarded from the encoding along with a few other tweaks described on the genealogy site.

The SOUNDEX code numbers (i.e. similar sounding):

1. = b, f, p, v
2. = c, g, j, k, q, s, x, z
3. = d, t
4. = l
5. = m, n
6. = r

And here’s an example:

Take the name “Carcassonne” and encode it:

1. Write the name “Carcassonne” on a scrap of paper.
2. Put a circle around the “C” because you always keep the first letter.
3. Cross out the vowels - “a”, “o” and “e”.
4. You’re left with Crcssnn.
5. Insert Soundex code numbers for these letters you arrive at: C-62255
6. We only count one of any two contiguous letters that are encoded the same so we end up with C-625.

Sounds Like Magic!

So we have encoded “Carcassonne” into C-625. Great! But wait, here’s the magic:

1. SOUNDEX(’carcussone’) = C-625
2. SOUNDEX(’carcuson’) = C-625
3. SOUNDEX(’carcsonn’) = C-625
4. SOUNDEX(’crason’) = C-625
5. SOUNDEX(’karkasone’) = K-625

Common mis-spellings 1-4 are also encoded the same! This means that we now have a mechanism to match up game names in our games list that sound the same, but could actually be spelled quite differently (e.g. number 4 above) thus making a great ’sounds like’ search facility.

Okay, so there’s limitations such as in the last example - although the words sound the same, because we keep the initial consonant, we don’t end up matching the two words, but it’s almost there!

Either way, its still better than returning a useless “no results found” message when searching for games so I’ll be incorporating this search trick into the games search module which I’m developing next.

Isn’t it amazing what some people dream up!

-Dan

Posted by Dan on Fri Mar 7th, 2008 at 5:54 pm and is filed under Development. You can follow any responses to this entry through the RSS feed. You can leave a response, or trackback from your own site.

• Posted on 7th Mar 2008 |
• Digg |
• Del.icio.us |
• Stumble