US Patent 10061766 Systems and methods for domain-specific machine-interpretation of input data

In an embodiment receiving input data over a computer network. The method also includes automatically generating topic-based metadata of the input data in relation to a multi-level taxonomy. The method also includes automatically determining parts of speech of tokens of the input data. The method also includes searching a metadata store using the tokens and the parts of speech, the searching including identifying, at a first level and a second level of the multi-level taxonomy, nodes that include metadata which sufficiently match the tokens and the parts of speech of the tokens, where the searching yields best-match grammar-based metadata. The method of also includes revising the best-match grammar-based metadata based at least in part on the automatically generated topic-based metadata.