PatentClassifier Has Arrived!

PatentClassifier, a new, free patent classification search tool, has been launched. The search tool is designed to suggest the patent classes closest to the phrases or words used to describe a particular patent. Get a low down on it after the jump!

What is it?

PatentClassifier is a new tool that will allow you to easily discover the correct class for your prior art, patent clearance and freedom to operate searches. Using the correct class during the prior art search and filing the application is of critical importance when you want the prior art to be thoroughly examined. The large number of patents filed every year makes it tricky to find the right class for a particular invention. Most times, different classes apply to a single invention, making it difficult to find the right set of classes. This is where PatentClassifier comes in. Using modern text classification, semantic search methods and data mining, the free tool comes up with the most appropriate classes pertaining to the invention.


How Does it Work?

The PatentClassifier tool is pretty straightforward. All you need to do is enter phrases or paragraphs that describe your invention and click on the Classify button.

PatentClassifier Interface

PatentClassifier Interface


The search results appear on the same page right beneath the Classify button. The results are displaced in a minimalistic fashion with the patent class listed in the first column followed by the class name. The third column of the search results is a Show Examples button; clicking on the button will display a list of patents that have been filed under that class along with a column displaying the subclass and a link to the USPTO website for that particular patent.

PatentClassifier Search Results

PatentClassifier Search Results


PatentClassifier claims a high accuracy rate for coming up with the appropriate patent classes. The company explains the Cross Validation methodology behind the accuracy rate thus:

 We used the holdout cross-validation strategy to measure the accuracy of our classifier. Specifically, we divided the set of U.S. patent grants from 2001-2014 into two groups – the main training set with 2,179,828 documents and the validation set with 42,414 documents. For each document in the validation set we predicted a ranked list of possible patent classes. In the ideal case, the right class would be always the first element in the list. Given that there are >400 main classes in the USPTO classification, we consider prediction to be correct if the right class is in the top 5 predicted classes and acceptable if the right class in top 10. We were pleased to find out that among the validation samples, 84% were predicted correctly (i.e. the right class among the top 5 possible classes) and 91% had the right class in the top 10 predicted classes.


In addition to the patent class search system, PatentClassifier offers tools like the USPTO Tree View and the Keywords search tool. As the name suggests, the USPTO Tree View will auto-suggest the classes to which the term or class number you have typed belongs. The Keywords tool on the other hand will allow you to search for classes based on whether they contain the keyword you are searching for. Both tools are simple to use with a clean, clutter-free interface.


Take the PatentClassifier out for a spin and tell us about it in the comment section below!
Patent Searches from Landon IP

This post was contributed by Abhishek Tiwari. The Intellogist blog is provided for free by Intellogist’s parent company Landon IP, a major provider of patent searches, trademark searches, technical translations, and information retrieval services.


One Response

  1. I did nuclear fusion stellarator and the answers were junk!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: