as we know, there is a long but unsuccessful history of attempts to parse, "understand", and summarize unstructured text. ditto the worlds of ontology, folksonomy, tagging and metadata schemas. ditto again the world of semantic nets, concept mapping, brainstorm visualization, etc. given the clear limitations and continued discontent, the core problem behind all of these remain unsolved: to create a structure which possesses properties that allow modeling, tracking, and comparing of texts.

to summarize what has been learned:

1. a single topic is insufficient — "table" is ambiguous unless in the presence of other topics. [table place-settings diners] is different from [table orthogonal-reflection vectors].

2. a base list is insufficient — lists may delimit components but do not express relationships.

3. expressing relationships among distinctions IS the meaning of a piece of prose — therefore, HOW the relationships are expressed is the key. semantic nets became mired in the nodes-and-links game and were never made to work.

i doubt that's the full list of requirements, but it does eliminate so many of the structures that have been proposed.

however, cybernetician Gordon Pask developed "entailment meshes" for the purpose of expressing understanding that results from from conversation. it seems to me that these are candidate structures for fulfilling the needs expressed in this post, and for the 'future of context'.