project idea: machine learning program that decrypts legalese to find similar patents
i wonder if uspo would be interested in licensing it if it was effective enough
i wonder if uspo would be interested in licensing it if it was effective enough
<Obvious-little-Idea> => <Converter> => <Awesome-sounding-invention>One of the jobs of the patent attorney is to claim the widest possible area that's supported by the invention disclosed in the application for a patent. You don't have a calculator, but "calculating means" which could encompass many different facilities - chemical, mechanical, electronic, quantum, ...
From the first example:
generating, using a processor = calculating
a plurality of images = images
clusters = groups
It's more of a translation software than machine learning.
Perhaps the translation software could suggest 3-4 variations of the legalese sentence and allow users to vote for the best translation?