In this issue I cover using natural language processing with “big data” for both overall topic relationships and auto-generated content suggestions for a given plot/location branch. You’ll find step-by-step tutorials for setting your system up for having these techniques available at your fingertips:
Discoverer’s Digest October Issue: New Horizons · Narrative Intelligence · Topic Modeling
This and other editions of Discoverer’s Digest are archived here:
ifarchive.org/indexes/if-arc … igest.html
I had read The Second Machine Age by Andrew McAfee and Erik Brynjolfsson, but it’s fascinating to see the concepts applied to build virtual worlds out of text.
When I read this month’s issue of Discoverer’s Digest, I was surprised by the amount of work that a single person (assisted by computer programs) can do; I’m not sure how much time that a team of people would have needed to do something similar just ten years ago.
My survey of today’s technology tells me that we’re not yet to a place where the systems can dynamically write prose and track the author’s objective.
This isn’t as bad as it sounds, I believe, because the techniques that I outline in this issue of DD lay the foundation for real-time systems. When GAI comes it will be a relatively straight-forward matter of bolting it to the architecture I describe.
Briefly I describe the results of this month’s DD issue here: you will have a method by which all your broad, interconnected IF topics are laid out using deep relationship modelling based on the entire work of Wikipedia (over 5 million articles).
Second, for each description in your IF you have a system that reaches out for object relationships in your prose for you to build on. A parts-of-speech tagger ensures that all facets of your IF is tagged for you to include appropriate responses to the player’s queries.
All these technologies are modular by design: if you create “glue” code you can imagine the day when you set the system with general parameters, let the system run for a few hours, and have it come back with automated diagrams of relationships and summarized/tagged prose of each result.
From here humans review the results and upload changes to a revision control system for the rest of the team.
I describe here massive IF world building systems. However, I targeted the articles to you, the individual author with an eye toward “straight forward” scaling them to massive proportions.
NOTE: this issue contains a lot of content and is designed for tablet screens or larger; it will take several moments to load.