Project

Patent Semantics

Groups

The automated processing of natural language by computer has become a paramount concern in nearly every field of human endeavor. The amount of information available in current literature and even the mass of information being generated on a daily basis is well beyond the scope of human understanding in any detail. Because of the sheer breadth of information, human analysis and curation of this information is rarely comprehensive and often error-prone. Machine-aided curation of this information could increase its utility by allowing more comprehensive analyses, more complete summarization, and extensive comparisons to current knowledge. Such curation requires a method for arriving at a high-level semantic description of the text at hand. Patent Semantics is an attempt to attack this problem in the domain of biochemical synthesis, by building structured representations of the procedures involved in different synthesis descriptions, grounding the components of those models, and developing algorithms by which detailed, human-readable comparisons of the descriptions may be produced.

The automated processing of natural language by computer has become a paramount concern in nearly every field of human endeavor. The amount of information available in current literature and even the mass of information being generated on a daily basis is well beyond the scope of human understanding in any detail. Because of the sheer breadth of information, human analysis and curation of this information is rarely comprehensive and often error-prone. Machine-aided curation of this information could increase its utility by allowing more comprehensive analyses, more complete summarization, and extensive comparisons to current knowledge. Such curation requires a method for arriving at a high-level semantic description of the text at hand. Patent Semantics is an attempt to attack this problem in the domain of biochemical synthesis, by building structured representations of the procedures involved in different synthesis descriptions, grounding the components of those models, and developing algorithms by which detailed, human-readable comparisons of the descriptions may be produced.