I've been wondering why we don't apply modern NLP techniques to the traditional parser-based interactive fiction format. In particular, part-of-speech tagging ought to be very robust by now, so we should be able to handle sentences with complex relative clauses and homonyms pretty well now, making the whole experience less frustrating.
Various levels of NLP do get applied in IF, but it doesn't necessarily help much. Being able to parse input is an important step, but the game/story needs to be able to do something with those verbs and nouns. For a good player experience the author needs to communicate clearly what range of input will be acceptable, making sure all nouns mentioned in text can be interacted, any actions implied in the text or by the nouns themselves are recognised, etc. If you do that well then you can have a good player experience with a purely handcoded parser.
You either hand some of this over to a ML model to handle, and lose some control / deterministic-ness, or you write a ton of code to handle the combinatorial explosion of potential parses of the sentence.
I guess AI Dungeon 2 sort of does this, though it's more like an advanced chatbot in the guise of a text-based adventure game, rather than an actual text-based adventure game.
usually a good-sounding idea like this, if it hasn't been tried, must have some hidden pitfalls. but the number of people writing parser-based interactive fiction is ludicrously small, so probably just nobody thought to do it.
