Using AI classifiers to automate patent landscapes
1st February 2019 |
The current approach to creating patent landscapes is broken. The dominant approach is Boolean search coupled with manual review, and this in not efficient. This two-step process is laborious and depends on having a searcher trained in patents, Boolean and the target technology. At a time when new technologies and innovation dominate corporate agendas, there is a pressing need to address this issue.
Harnessing AI to classify patents means that this process can be largely automated. By training a classifier, it is possible to achieve search results that are equivalent to patent landscapes built by humans. The solution applies a new set of algorithms that look at all relevant patent data, and are not restricted by human defined rules. Classifiers dispense with the need for keywords or over dependence on patent office codes that do not align with most companies’ views of technologies.
There are many compelling reasons why AI classifiers should be widely adopted:
- Machines don’t get tired: while analysing 100m patents is labour intensive for humans, computing power is a near infinite resource
- AI is fast, really fast: classification is perfect fit for AI. Let the machine find patents that map to a technology and use the extra time for added value analysis
- Trained by experts, accessible to everyone: AI classifiers use human curated training data (patents that are positive and negative examples of the technology in scope). Once defined classifiers can be used by anyone – no patent or technology expertise required.
- Eliminate the need for manual intervention: AI classifiers generate results which are right first time. This streamlines the process making essential information accessible at a much lower cost.
Put Cipher to the test. Contact us for a demonstration on firstname.lastname@example.org
Cipher is the only commercially available system to have fully automated AI patent classification. Cipher Automotive and Cipher n/d are taxonomies built for the automotive and technology sectors and each include 150+ pre-trained classifiers. Cipher subscribers have the ability to customise libraries to their own internal taxonomies, an approach that has been adopted by companies such as ARM, BAE Systems and Ocado.