I've been reading Josh's (your) devlogs recently about the development of a "newsfeed" that sort of tells the story of each Limit Theory game (at least that's how I've been understanding this project), and I've had a couple ideas that I'd like to share.
The first idea would be to take a large dictionary of words and map each of them to the same attribute space that Josh uses to describe the "personality" of AI "players" and colonies, and then sample that space appropriately when filling in syntactic patterns determined by each "Action" being described. This seems really memory intensive, and it would be a really interesting (hard) problem to accurately map each word.
A variation on that would be to use sub dictionaries depending on each event. For instance Wikipedia pages describing historic battles could be scraped for dictionaries to describe regional conflicts. It might still have the memory problems, but might add a more believable/recognizable tone...
In considering how words communicate tone, I was thinking about the "directness" of a word. Basically I mean a combination of tense and specificity. The more specific and the more immediate a word is, the more "aggressive" it sounds. Conversely, the more vague a word is, the weaker it is. If this idea were fleshed out a bit more, then maybe WordNet http://wordnet.princeton.edu/" could be used to efficiently find appropriate words.
Actually I think WordNet might be a good tool for this application anyway... or maybe constructing wordnets from relevant sub dictionaries.
Anyway... I'm a long time lurker, and finally felt I had some grain of sand to contribute.
Users browsing this forum: No registered users and 2 guests