Dev Blog: It’s all semantics

We’ve been looking into ways of measuring how to reward writers in the Written World, by thinking in terms of understanding the “quality” of the stories. So the question is - how do we take a ruler to success in story writing?

We first started looking at the most basic level - firstly checking if people have finished the story and played all their encounter cards. Then we started looking into other more subtle measures - 

• Checking the amount of words each users writes in a turn.
• Looking at The average typing speed.
• Looking at the type of language being used. 

Using this information we could infer a a level of “passion” for the story. We might later use this measure to work out how an optimum level of story engagement looks. Getting this kind of high quality data back as our users play is spectacularly interesting. Pulling together a series of quantitative data points like speed of writing and types of language used into a measure of the ‘passion’ of a writer lets us build game play objectives which support pulling a writer towards the point at which they are most engaged by their task. This is exciting stuff - we’re building an index which measures user engagement and learning how to keep people consistently in a state of flow.

Next we begin to look at the actual quality of the text itself. Using the fantastic AfterTheDeadline plugin, we can do some interesting heavy lifting on the text. We can reward correct spelling, and correct uses of grammar - going as far as penalising misused parts of speech (a/an, there/their etc), repeated words, and verbs in the wrong tense.  We can also do some more advanced checking of the tone of the speech, reducing complex phrases down to more simple ones,  and even removing redundant phrases. 

For the alpha, we’ll avoid using some of the more complex features available to us. I’m a bit worried about penalising users for mistakes without showing them why we’ve done it in the first place, so that’s one of the things we’ll address during play testing. Perhaps we will allow users to go back and make corrections to the text after it’s been written inside the flow of the story, or allow them to alter the text before it’s sent out.  Perhaps we’ll have a huge red box that appears and shouts at the user if they use too many cliches. I DON’T KNOW!!

Now let’s say we have a “passion” and a “style” index, which will allow us to measure and reward player actions in a rich way, from here we can think about how to understand each story within the context of other stories.  Using a semantic text engine, such as OpenCalais, we can start to make some incredibly rich connections to entities mentioned within the text. If you mention a location in your story, we’ll know which one and where it is in the Written World. If you write about pre-existing characters we’ll be able to tell you where else they’ve appeared, and in what way. 

If you’d like to see a demo, try copying/pasting a chapter from your favourite story here You’ll see a list of characters, locations, organisations mentioned and lots of other interesting data - We’ll be able to use the information gleaned in this way for some fascinating stuff. I put in a chapter from 1984 to see what information I could get out of it - it came out with a comprehensive list of the characters, and organisations such as the “Eastasian Government” - This is incredibly powerful because putting all the stories created through the Written World through this kind of processing, would allow us to create suggestions for themes already written about by other writers.  We want the Written World to be a  kind of “unified story”, where entities can grow in a rich and interesting manner, and we are confident with these kinds of tools we’ll be able to create it.