Guest Article: Behind the Scenes – Localizing a Word Game

This guest article was authored by Matt Decker, founder of Artifex Consultants. Matt has over a decade of experience in localization and translation. After his last position as casual games Localization Manager for an international publisher, Matt turned his attention to assisting casual games developers, among others, in reaching international markets in a more cost effective and streamlined manner. Drawing on his extensive contacts with language professionals and service providers like ENLASO, Matt aims to make that process as painless as possible. Contact Matt at: mdecker@artifexconsultants.com.

Games are supposed to be fun to play...
Games, by their nature, are supposed to be fun to play, n’est-ce pas? That is, most games are as long as they don’t involve driving your opponent into bankruptcy (namely you being the unlucky opponent) or owning large swathes of the known world (or you with the ever-shrinking share of the known world).

When I was the localization manager at a large international casual gaming publisher, one of the very first “casual gaming” localization projects I ran concerned “Flip Words,” a popular English downloadable word game. Upon first blush, localizing the game appeared to be relatively easy. This was deceiving. As we dug further into the game and game mechanics, numerous issues presented themselves. This article summarizes some of these major issues and our solutions.

Play of the game
Flip Words’ rules are rather simple. The player can:

• Form a word using any combination of letters in a chain. Words can be no shorter than three characters or no longer than twelve. AND, they must be adjacent and linkable to other letters on the board to form the word.
• Guess a mystery phrase at the top of the screen. As you form words, the first letter of each word is revealed in the phrase. Guess the phrase to win the round and move on to the next puzzle.

Challenges and solutions
After analyzing the game from the localization perspective, the team moved on to a more thorough analysis of the game code and the actual mechanics of the game. Sacrifice the play too much in the interests of cost and schedule and the game would lose all aspects of playability for the target audiences and would perform poorly (the target languages were French, Italian, German and European Spanish). Those issues that would affect the game were:

• Letter distribution,
• Localized dictionary,
• Phrases within the game and finally,
• Accounting for vowels and extended characters.

Challenge 1: Letter distribution
The letters appear on a game grid of 7x7 tiles. On the grid, the player creates words by linking letters in order. As stated above, only words from 3 to 12 letters are valid and the more letters you use, the higher your score. The key issue here is that the frequency and distribution of letters were tailored for English. For example, certain English letters are more apt to appear than others. This is also valid for English-centric letter combinations.

Solution:
A solid solution here was critical; if we misjudged the requirements and didn’t present the player with the ability to spell words easily in their language, the game would be unplayable. After identifying the most appropriate distributions and frequency patterns for each language, the team proceeded through a series of modifications to the algorithm and linguistic testing to ensure that we could get the best mix and game play possible. Each language required its own specific modifications and therefore its own unique build.

Challenge 2: Localized dictionary files
The English source dictionary contained a word list of approximately 110,000 English words. The game engine would cross-reference words input by the player against the entries in the dictionary and would validate word submissions by the player at runtime and award points based on the letters used (the more infrequently used letters and letter combinations submitted, the more points be scored). Additionally, as stated above, the initial letter in the word would reveal a letter in the phrase.

Solution:
Obviously, the key here was to create a comprehensive dictionary for each target language. Word lists exist in the public domain and, in some instances; rights to the game are free and only require that you credit the owner of the word list. As crediting the owner of the list was not an option for this game, fresh word lists were created for each target language. The greater challenge here was to review the lists to ensure that no inappropriate content was included (vulgarities, questionable slang, etc). On the plus side, the target languages have a richer base from which to draw as compared to English – verb conjugations and adjectival declensions.

Challenge 3: Phrases used within the game
The stock English phrases shipped with the game specifically target the American market. There are several genres from which the game pulls the phrases for each round of play. Additionally, functionality within the game allows players to submit phrases to a server. Submitted phrases are reviewed for appropriateness and, if not flagged as inappropriate, are released to the gaming community (the game system requests to contact the server and download new phrases). Average submissions for English were approximately 500 phrases a month.

Solution:
Phrases that were too American-centric were discarded automatically. Linguists assigned to each language would then create a database of new phrases based on regional popular culture. The localized versions also support phrase submissions. Native language editors are assigned to review the phrases prior release. Phrase submissions provide an actual benchmark as to how well the game has been received by the public. Of the four languages, French proved the most popular and average monthly phrase submission approximated 300 – 400 phrases and still sells 50 – 100 copies a month on one site (excellent performance for a title that is over 18 months old).

Challenge 4: Accented characters
In a total localized solution, accented characters would appear in three locations and each location would require a different solution. Characters appear on the game board, are displayed above the game board as the last word correctly spelled, and finally in the phrase itself.

Solution:
The characters on the game board are actually image files (each letter has three associated images) a graphical solution was too expensive and time consuming. Given that, the decision was made not to differentiate between accented and non-accented vowels for the game on the game board itself. The entries in the dictionary files would still retain the accented characters and the extended characters would show up on the screen as the word was formed and in the phrases themselves. Behind the scenes, game code was modified to map extended characters to their counterparts. Thus: à would map to a, é and è would map to e, etc. This approach allowed the player to spell “etre” on the board, whereas the dictionary would recognize it as a valid entry for “être”.

Conclusion:
As far as casual games for localization are concerned, the Flip Words project presented many new exciting challenges and wrinkles. Released in early 2006 to online game portals in France, Germany, Spain and Italy, the localized download version of the game drew immediate attention. To date, the title continues to sell well on several major French sites, with average download totals approaching 100 – 150 instances a month. As they say, “Not bad for an older title.” In 2007, the Flash version of the game was released in the same target languages. ENLASO was a key resource in localizing and engineering unique solutions to this version of the game.

ENLASO's Games Localization Solutions
For more information on how ENLASO can assist you with all of your localization needs, please contact Chris Raulf at craulf@translate.com, call 866-415-6820, or complete the quote request form.