Semantic and deep learning based prior art query for abstractive patent summarization

Abstract

Patents are an excellent source of information on inventions. They are valuable because of the technicality within the document, including drawings that are not even represented in the scientific literature. At the same time, they are essential and economically highly relevant legal documents. They are one type of Intellectual property that provides exclusive rights for the Novel, Non-obvious and Useful information. It has all the details regarding the invention, displays the technology trends and says what the competitor is looking into. When there is an invention, the work of all researchers and all patent searchers worldwide revolves around the need to know precisely what has already been done and published before. There comes the need for patent search or, more prominently, Prior-art search. newlineOver recent years the volume of filings and related data has dramatically increased, as the capacity of IT systems to store, search, and process this material. At the same time, the growth of the Internet has changed public expectations and, indeed, the needs of examiners and the public, particularly prior art searching. The patent documents written by the patentees have their lexicons to describe their inventive details. They often include different data types, drawings, mathematical formulas, bio-sequence listings, or chemical structures requiring specific techniques for effective search and analysis. The classification code assigned by the patent office assists in managing their examination workload and searching patents, but these classification codes are not harmonized across different patenting offices. newline