Introductory video: watch it here.
Unique algorithms (classifiers) can be created and trained to find specific patents in key technology areas of interest. To determine and “learn” your definition for the technology of interest, the software evaluates claims, abstracts, keywords, citations, class-codes etc. from the patent metadata. This is different to usual forms of searching which rely on keyword and semantic searching.
The classifier can take anywhere from 1 hour to 1 day to build, depending on your subject-area knowledge and how broad or how narrow you would like the search to encompass.
I.e. If you wish to identify all patents related to your definition of ‘electric toothbrushes’, you will need to teach the machine using examples that fall within and outside of the technology you're defining.
Providing examples of what you are (electric hand-held device for cleaning teeth) and are not interested in (e.g. medical instruments used for oral surgery) helps the machine ‘learn’ to include/exclude them from the search. The benefit is that you don’t have to spend time thinking of all possible keywords and synonyms that could be used to describe this iteration of the technology.
Once built, the algorithm (Cipher Classifier) sits in your account library and is available to use on demand. This means that it can be used repeatedly for any number of reports that you wish to generate.
Build Classifier > Run against patent portfolios of interest > Instant insights.
Cipher Classifiers are faster than conventional search methods to build and run, can be used repeatedly across report variations, and constantly improve with all refinements.