Language Processing

Those of you who played text-input adventure games back in the day (King’s Quest, Leisure Suit Larry, etc) will recall how fun it could be to test the limits of the game designer’s imagination by experimenting with language commands. It was thrilling when you tried something “unusual” or “outrageous” (in your mind) and yet the game responded appropriately. Of course, it was also frustrating when you tried to accomplish something serious and the game didn’t understand you. (For an exercise in said frustration, give the much-hyped Facade a try if you haven’t already. It’s a glimpse into what made these games fun, and everything that made them less-than-fun.)

At any rate, most of that “joy of experimentation” disappeared when adventure games migrated to mouse-only. Perhaps not coincidentally, adventures games themselves began to disappear soon afterwards. But text-input has returned in the form of viral marketing gimmicks like the Subservient Chicken campaign, and in IM bots like Spleak, which capture the imagination in part by encouraging users to test the limits of the designer’s vision and resources via text input. Both the Subservient Chicken campaign and Spleak have proven quite successful within a limited but significant audience.

What does this teach us about how language processing can be reabsorbed back into the world of games? What light can it shed on the strengths and weaknesses of a game like Facade?

Here’s one thought. Yesterday, you published an adventure game and were finished with it. Today, you can monitor user queries, capture the most common (unhandled) queries and create new content on the fly to address them. The game never needs to end (and indeed, neither Subservient Chicken nor Spleak appear to have a terminus)… as long as the game isn’t defined in a linear fashion. In fact, it’s not even clear that the game needs to be limited by the designer’s time and resources; what about taking advantage of crowdsourcing in some way to increase the “intelligence” of the game engine?

Gotta run now, but I’m curious to hear what thoughts & ideas you might have!

7 Responses to Language Processing

  1. Yes, but thats just step one. What you really need is a robust sort of inference engine in order to contexualize that data in a way that the systemic attractors of the crowd/AI/processor arcitecture are balanced and symmetrical. For example, idioms are very difficult to disseminate without producing confusing permuations that actually improve the system, rather thany muddle it. Speech is extremely heterogenous, there is no such thing as standard grammar or venacular when dealing with any kind of diverse audience.

    But there is hope, for example: http://www.novamente.net/

    Love to know what you think of that, btw.

  2. Heh, for some reason I read this and start wondering if I’m getting a free copy of Zork with the Chatpad…

  3. @Patrick — hadn’t heard of Novamente. For some reason, their videos won’t play on my PC. Sounds like what they do is pretty interesting, but hard to know from what little I’ve seen how much is hype (or not.)

  4. We lost a lot when we stopped building on text parser technology. We know have shallow story-telling in games, we have stilted conversations trees and boring multiple choice decision making. More than anything in video gaming today, lack of story-telling in games is leading to it’s eventual demise. Without better stories gamers will ultimately get bored, regardless of graphics or sound or animation improvements, because all these hang on the story.

    Produce a conversational text parser and a whole new range of games could be released that had better stories, better world creation and better character communications and item manipulations.

Leave a Reply