Decomposing documents into claims

Turning a large corpus of internet discourse, academic papers, and the like into a clean database of claims is the foundational subproblem. A document must be read, the propositions it asserts must be extracted, and each one organized into the graph. Several distinct difficulties fall out of this:

• Splitting vs. lumping similar claims
• Canonical forms of claims
• Structuring the claim graph (Wiki vs. DAG vs. DCG)
• Defining edges between claims
• Determining claim importance

Once a claim has been extracted, it still has to be reconciled against what is already there — see matching instances of claims to claims in the graph.