squirrel VS plint

I wrote a static analyzer for patent claims: https://github.com/btrettel/plint

There are tons of ambiguities and other issues in issued patents!

When I get the time, I intend to rewrite a significant amount of the code in an object-oriented style. (Basically, making a class for a patent claim and having a ton of methods for that.) The code right now is a mostly untested mess that seems to produce decent results but is not pleasant to work with.

I'm a patent examiner and I can't agree with "It's very precise if you can learn to read it". A large fraction of patent applications I've worked on are unclear. I even wrote a linter to help catch some problems: https://github.com/btrettel/plint

(DOC lawyers make me add this: This post is just my personal opinion, not that of the USPTO, DOC, US govt., etc.)

The USPTO uses a lot of automation, but mostly not where they should.

Note: I am a current USPTO patent examiner, and this is my opinion, not that of the USPTO or US govt.

The USPTO apparently has two contractors to classify patent documents. I've heard that some sort of AI system is used for classification, in combination with a lot of poorly paid contractors. In my experience, the classification is so frequently wrong that this is clearly not working. It might seem okay to upper management, who never has to actually deal with the classification being inaccurate. But examiners aren't happy with it.

Many people are calling for AI search. The new head of the USPTO mentioned it during a recent all-hands meeting. Unfortunately, the people who propose AI search don't seem to realize that 1. the USPTO has at least 5 AI search tools at their disposal (PLUS, More Like This, Dialog's similarity search, IP.com's similarity search, and Google Patents similar documents) and 2. none of these AI search tools work that well. In my experience, most of the time these tools don't return useful documents. (I still try them for every application as there's little downside.) The documents are usually close but it's rare that I'll actually use one of these documents in a prior art rejection. AI search sounds good to people who have never searched for patents and particularly have never used the existing AI search tools. AI search technology probably won't be good for a decade or more.

In contrast, tools to analyze patent claims for various problems (basically, linters) have been available for around 30 years and can be quite useful in my opinion. But the USPTO has no such tool available to examiners, and analysis under 112(b), etc., is almost always done manually. I wrote my own tool, which I run on my USPTO computer on a regular basis. You can see it here: https://github.com/btrettel/plint

There are a huge number of opportunities to streamline USPTO operations with automation. Why are IDS forms not computer-readable? Why is so much information not auto-filled? Why do I have to fill out "bib data sheets" for every application? Why do I have to manually upload my search history when the system could easily automatically grab it for me? Etc.

And automation isn't enough. There should be more data validation in the process, as a lot of problems can be automatically caught at the time of filing or when I post an office action. That's when fixing these problems would be easiest.