[garfieldnate]

Lots of linguists only work on static structures, diagramming morphosyntax and semantics of a standalone utterance. All of the fuzzy social context stuff doesn't generally get broken down into a "mathematical" structure in the same way. The piece we're missing here is called "grounding", the mapping of words in the utterance to entities in the context. I think it gets largely ignored because of this study of static structures. I don't know of any generic parsing framework/theory/tool that comes with grounding out-of-the box. It's just not done. Please prove me wrong if you know of a tool that does this. Other keywords are "deictic" and "pragmatics". Generally you only start to worry about grounding when you're doing some kind of robotics or other human-computer interaction and you need a human's words to map to sensor data or something on the screen to understand what the intention is.

As a possible exception, anaphora/cataphora resolution is pretty well-studied and for example is supported in spaCy (https://spacy.io/universe/project/neuralcoref), but this is mapping one word in an utterance to another word in the utterance, rather than to an entity in the context.