How patent examiners can quickly find new prior art – case study on Nike shoe patent

In my time with Ambercite I have been lucky enough to meet many patent examiners, and in this process I have become aware of how diligent the patent examination process can be – in many cases taking two to four days to search for relevant prior art.

These days patents that are to be examined often come with a pre-existing list of prior art, for example prepared during the PCT examination phase. As an example of this, consider this Nike patent WO2014143811, for a Shoe with resilient heel. If we look up this document in Google Patent or Espacenet, we can see that it has four listed citations:

  • US5955957, Footwear with electroluminescent wire
  • US7100308, Footwear with a heel plate assembly
  • US7789520, Electroluminescent communication system between articles of apparel and the like
  • US20120192456, Shoe With Resilient Heel

In the International Search Report the second of these patents is regarded as a X citation (relevant to novelty) while the other three patents are regarded as Y citations (relevant to inventive step).

A patent examiner asked to examine a national phase application for this WO patent could simply rely on this existing search report, or wish to expand the search to see if there are any relevant patents that the PCT patent examiner might have missed. As discussed, search reports can take many hours or days to prepare – but there is a very fast and simple means of finding new prior art.

This is the process of Cluster Searching, recently introduced by Ambercite. Cluster Searching identifies the most similar patents to one or more starting patents based on similarity in their citation profiles. By doing so, the time needed to prepare what could be quite erroneous keyword or classification code searches are avoided, allowing the searcher or examiner to focus on the results instead.

Cluster Searching only has three inputs: The patent numbers, a simple date filter, and the number of results desired (up to a maximum of 2000). For the case of the Nike patent, we might ask to only show similar patents filed before its priority date of March 2013.

But what starting patents should we use? The patent number itself is an obvious starting patent. To this we might as well add the known prior art patents – as otherwise Cluster Searching would simply be telling us what we already know.


Note from the above query that three of the patents have ‘0.5’ written after them (their ‘weighting’ value). This is to discount their influence by 50% compared to the Nike WO patent and the identified X citation, which, unless specified, are given a default weighting of 1.

We ran this query by hitting the big orange search button.  The following result is shown after just a couple of seconds and the screengrab below shows the top 20 results, out the 100 we requested. These are ranked in order of overall similarity to the starting patents, so we would expect the top listed patents to be the most similar to the Nike patent and its prior art.



While in this table only the patent titles are shown, each of the little blue circles in the fifth column links to more details about each patent, for example, the patent abstract. Clicking the orange circles in the sixth column opens up the patent in AmberScope, our graphic based patent search engine. In addition results can be sorted or filter by any of the column titles, for example to exclude patents owned by Nike.

100 patents can sound like a lot to review, however they are sorted in order of likely similarity. This means that an examiner or searcher can spend more time reviewing the patents at the top of the list, and less time reviewing the patents at the bottom of the list, thereby making smart use of his or her time.

It is also possible to copy any of these fields, say the patent number, into any other patent database, or to export this table as an Excel spreadsheet.

One of the most important things to note, and a capability unique to Ambercite, is the identification similar, but “Indirectly” linked results. The second last colum on the right makes a distinction between a ‘Direct’ citation (a listed patent citation to any of the starting patents) or ‘Indirect’ citation, which is only indirectly connected to the starting patents. This ability to quickly go beyond known citations is one of the unique and key features of Cluster Searching – and is further discussed in this article published by Tony Trippe.


Do the results help invalidate the Nike patent?

This is for others to decide  – but a simple review of these patent titles suggests that many are in very much the same area as the Nike patent. A careful examiner or opponent trying to invalidate this patent would be well advised to review these patents – and can do so without spending hours preparing search queries.


Can patent searching really be this simple?

To patent searchers or examiners who are used to spending hours carefully crafting queries that can produce a lot of irrelevant hits and still miss relevant patents, this can seem almost too easy. But, the process is easy thanks to their examiner colleagues. Cluster Searching is building on the hard work already completed by examiner colleagues and applicants alike, as Cluster Searching mathematically combines all of these citation opinions in the landscape. Because of the statistical nature of our algorithms and the high number of data points, individual errors made by examiners or applicants are being softend. Sometimes life can be easy.

Cluster Searching can provide real benefits in productivity compared to more conventional searching. This can speed up the patent examination process and reduce examination backloads.


What do patent examiners say?

We asked a friendly patent examiner working in a major patent office to review the validity of our results. He gave us a list of prior art references he had already found after an extensive search on a patent he was examining. We then provided him with the list of ranked results and asked for his feedback – which was:

“[Top ranked patent] was very good because it found an element that is hard to search for, i.e. ‘independent power settings’.  The ‘independent power settings’ are hard to search for because searching for something like “power near2 settings” would return several thousand hits and most of them would not be for independent power settings and limiting that search with an adjective like “independent” would filter out a lot of good results.  And it’s amazing that it came up first in your search…”

   [note that we have changed one of the technical terms to preserve the confidentiality of this patent application]


Can I try Cluster Searching for myself?

Ambercite is offering free trials to selected applicants. Please contact us for further details – and understand for yourself just how simple and fast patent seaching can be.

Read More