Uptick in unsupervised machine learning to manage and analyse patent data

A patent is a valuable asset for a company or individual – any missed deadline could result in irrevocable loss or a hefty fee to recover it. Various patent professionals work with patent analytics teams to analyse and manage patent information. Currently, these files are managed manually, which involves the use of office tools (eg, Word or Excel). However, with the astounding growth of intellectual property throughout the world and the steep increase in the number of patents granted every year, patent professionals are facing myriad challenges. Analysing patent information manually is highly demanding on time and resources too, as files can sometimes run in to thousands of documents.

With growing competition in various industries, companies need to know their competitors and their areas of interest. However, it can be challenging to allocate appropriate monetary and human resources to analyse the patent information.

To tackle this, law firms have adopted various patent management tools based on machine learning. In particular, unsupervised machine learning tools are being used to manage patent data and big data in many areas of manufacturing and operations. This has proven to be greatly advantageous for low to mid-size companies as these tools provide them with the high-quality data necessary for them to survive in a competitive marketplace.

The following patent activities can be managed using unsupervised machine learning:

  • training a system based on unlabelled and/or uncategorised patent data;
  • identifying the knowledge focus and dynamics of a particular technology or domain; and
  • discovering various trends, patterns and relationships between unlabelled/labelled patent data.

Unsupervised machine learning tools eliminate the need for human effort since they produce an outcome based on input alone and require no feedback. Further, these tools are capable of managing and analysing patent information by directly connecting with patent offices. Search analysis can be performed based on the full description of a patent application and not just abstracts, claims or the application summary.

One example of an unsupervised machine learning tool is the Knowing, Mapping and Exploration (KMX) tool, which allows for:

  • the rapid classification of a large number of patents or scientific articles based on very minimal training examples, which provides versatile and agile analysis; and
  • a very swift full patent analysis by ranking patents based on key features of the invention.

The advantages of unsupervised machine learning for managing patent data are as follows:

  • it is time and cost efficient;
  • it provides high-quality analysis;
  • there is a higher success rate during the patent examination stage; and
  • it creates a single platform for all jurisdictions.


Rather than performing tasks such as analysing the full description of an application, classification, tracking upcoming responses to office actions and payment of maintenance fees, among other things, use of these automation tools streamlines tasks and allows patent professionals to concentrate on more substantive work.

This is an Insight article, written by a selected partner as part of IAM's co-published content. Read more on Insight

Unlock unlimited access to all IAM content