A question we often get is ‘How does Cluster Searching work?”
Without giving away too much of the propriatory details, Cluster Searching uses citation links to identify similar patents, Once people know this, they often ask something along the lines ‘does this use keyword matching as well?
To which our answer is “no – there is no direct use of keyword matching to identify similar patents.”
Some people find this answer quite surprising as the keyword searching is used as the basis of the majority of patent searching.
But in fact, citation searching does not need keyword searching. The reason why is that every citaiton link is an opinion by a specialist that two patents are similar even if they use different keywords.
As an example, consider Nike patent US7945343, filed by Nike in 2006 for a Method of Making an Article of Footwear (via its design in a graphical user interface)
Perhaps more interesting, these 26 backward citations have all been selected for their relevancy. If you were to simply run a relative simple search query for say all patents listing (shoe OR footwear) AND (computer OR website OR online) in any of title, abstract or claims, you would end up with thousands of hits – I got 3567 patents when I ran this query in Patentlens just now.
So, which would you rather do – sort through 26 patents selected for their relevancy or 3567 patents found by a keyword search?
Even better, Cluster Searching does not list the citations, but uses proprietory algorithms to rank these citations to predict the most similar patents. So in the case of the Nike patent, Cluster Searching predicted the most similar backward citation to be US20050071242, filed by Nike for a Method and system for custom-manufacturing footwear
Simillarly, we can predict that the least most similar listed backward citatoin to be WO2000036943, filed by Reebook for An article of footwear and method for making the same, and which does feature the use of onlilne customisation at all – instead only factory customisation (hence the citation reference)
Even better, Cluster Searching can also identify indirectly connected patents that may also be relevant, such as US5339252 filed in 1993 for an Integrated system for foot measurement, last and footwear manufacture.
a data processing device for receiving the foot sizing data from the foot sizing device and for transmitting the foot sizing and footwear style data, and a computer automated design mechanism for receiving the data from the data processing device and deriving machine readable data in a form suitable for transmission to footwear last production machinery for manufacture of a footwear
This was not cited by either the examiner or the applicant as prior art for the Nike patent, yet appears to be relevant.
This is Cluster Searching in a nutshell – and shows how it is able to find relevant patent without direct knowledge of relevant keywords. Instead the accrued knowledge embedded in the citation links are able to both rank the most similar patents, and identify similar patents not cited by the applicant or examiner.
Want to try this for yourself?
Cluster Searching is a very fast and easy to use web application. Free demonstrations and confidential trials are available to qualified applicants – please contact us to arrange a short demonstration and trial.