Category: Amberblog

Apple receives a patent for screen control by eyeball tracking. What additional prior art can be found by an iterative Cluster Search?

Apple have recently recieved a patent US9189064 for control of a tablet by eyeball tracking. According to Business Insider, ‘this refers to your iPhone tracking where on the screen you’re looking — and delaying notifications until your gaze is fixed in the appropriate place.

This is an interesting patent, and also a chance to test the Cluster Searching patent search tool developed by Ambercite. Since the patent has been granted, we now have an independently created search report by the USPTO, found here

Cluster Searching works from a series of starting patents. In a lot of cases, searchers such as patent examiners might be starting from a search report for the PCT version of this patent – which in this case is WO2014039449.  This patent is published by WIPO with 7 patent citations, 4 of them supplied by Apple and more concerned with touch screens than eye control.

If we wanted to search for further prior art based on the WO patent alone, we might run a search based on the patent itself, and perhaps the examiner citations. We might put the rest of the cited art in the ‘hide from result box. And we would ask for patents filed before the September 2013 filing date of the WO patent, and ask for perhaps the 100 most similar patents.

The resulting query would look like this



The results would look like this, with . Of interest, the vast majority of the patents found are not directly linked to eye control of touch screens. but these ‘false positives’ are a fact of life in any patent search. Of these first five patents shown, the first two are related to eyeball control – and the other three are not. Normally our searches return a lower ‘false positive’ rate than this, but every search is different.



But what else is there? Patent titles in this case can be quite helpful, and so we scrolled through the results.Besides the first two mentioned patents, the patent ranked in 44th place was also interesting, being US7561143 Using gaze actions to interact with a display, filed by University of Arts (Philadelphia, US). This discloses:

The eye movement analyzer analyzes the eye movements for a sequence of gaze movements that indicate a gaze action which specifies an operation on the display.



Using iterative searching to expand your results set

We can further extend the search by copying the first two patent numbers, and the Gaze patent, in the search box. We can copy every other result we initially found into the “Enter patent numbers to hide box“, and so they will not be displayed again.

The search query now looks like this (note that only the top few ‘hide’ patents are shown):




The new results are shown below. Again not all patents are relevant, but many are – in fact, of these top 100 patents, 27 refer to eyeball tracking. 

We could stop here – or run one more iterative search by copying these 27 addtional into the Search box, and the reminder of these 100 patents into the hide box. This time the results look like this:


The top 5 patents all refer to gaze tracking, and so the majority of the other patents – for example US20110175932, Eye Tracker Based Contextual Action.


So as you can, we are working from just 4 patents disclosed we are rapidly building up a databset of relevant patents.  In fact, if we look at the results for all iterations, we have a database of 60 relevant patents in total


But which of these patents did the search reporf for US9189064 find?

Out of the 60 patents found by this three stages iterative approach, just 4 were also found in the USPTO search. And we have seen these divegence from conventional search before.  So by using a completely approach to conventional search, we end up with a completely different set of results – and if you are trying to produce the best possible search report, isn’t this what you want to see?


Want to try Cluster Searching 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. You may be surprised (and impressed) by just how much time and money you may save.

After the trial has finished, Cluster Searching is available to companies, organisations and individuals at a very competitive annual subscription – we are happy to discuss this in person.


Read More

Cluster searching is now even faster and more efficient – Case study of new ‘ignore’ function

Since its release in 2015 Ambercite’s Cluster Searching tool has picked up a wide range of users around the world who are impressed by its ability to quickly find patents missed by conventional searching processes.

These users have also provided some very valuable feedback to Ambercite, and part of this feedback is that these users sometimes end up seeing patents they have already seen in previous searches. This can slow down the reviewing of the results that Cluster Searching provides,  and can be inefficient.

It is of course impossible to know what patents that our users have already seen – but they will probably know this themselves. For this reason Ambercite has just upgraded Cluster Searching with a new feature that allows users to ignore previously seen results from the results that Cluster Searching shows. This operates through a second data entry box, shown below.


Patent numbers pasted into this box are ignored in our display of similar patents. It is possible to add up to 2000 patents into this box, and sometimes more in some cases.

Case study

Imagine you were asked to run a patentability search for a mirror for a bicycle helmet.

In this case we have no starting patents, but a keyword search in a more traditional patent search system for ((bicycle near helmet) and mirror) brought up 25 results. We reviewed these results, and found that the 13 of these patents were relevant to the invention. The other 12 were not relevant.

But have we missed any relevant patents? What else would Cluster Searching show us?

This is an easy question to answer. We simply enter the previously found but irrelevant patents into the ‘ignore’ box, and enter the five patents that we like into the normal data entry box. In this case we request the 50 most similar patents. The resulting query looks like this.


The results look like this.


By definition, we have seen none of these patents before. These for example, include US5432960, filed for a Helmet mirror attachment, and found in 3rd place on our ranked list of results.


Yet this patent was missed by the first patent search. Why? This particular query was looking for a combination of bicycle near helmet and mirror, but in the patent the words helmet and bicycle were not found closely together. OK this reflects a flawed search strategy (in this very simple case study), but no search strategy is perfect, no matter how sophisticated – in the absence of hindsight that is.

In contrast Cluster Searching relies on the expertise and search strategies of all of the search reports conducted by the examiners in the field (as expressed by citation opinions) – so you are not reliant on just one search strategy. This greatly increases the chance of finding the patents you are looking for.

This is a perfect illustration of the value of Cluster Searching – which has now been enhanced by the ability to ignore previously found results.


Want to try Cluster Searching 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. You may be surprised (and impressed) by just how much time and money you may save.

Read More

A simple case study of how Cluster Searching can find patents missed by conventional searching

A client recently asked ‘can you provide some examples where Cluster searching efficiently identified relevant citations, which were not identified using existing tools?’.

This is a great question. I was running a demonstration case the other day, and this was a perfect examples.

Being a demonstration case, we wanted to a simple invention, and this turned out be the concept of a surfboard carrier for a bicycle (Ambercite founder Doris Spielthenner is a keen surfer and cyclist).



The search process

1) Using a widely used conventional search engine, we ran the following search query.

Search for, within the title or abstract=((bicycle* or bike or cycle) AND (surfboard*)) – and with a CPC cclass code of B62J*, where this class code means accessories peculiar to cycles including article carriers. 

This produced 28 patent families, of which we judged 21 to be highly relevant, i.e specifically disclosed a surfboard carrier for a bicycle.  While this is an excellent relevancy rate, most searchers would agree that this much better than normal.

2) We used these 21 patents families as the basis for a Cluster Search, and produced a list of the 50 most similar patents to these 21.

3) We then cross-checked these 50 patents against the 21 families  to check for duplicates, and the remaining patents assessed how many of these specifically disclosed surfboard carriers for bicycles.  There were 7 such patents.

So this was 7 patents which where not picked up by a conventional search. This is great of course, and justifies the value of cluster searching, but the real question is why?

This is best explained by the table below.

What additional patents did we find? 


Why was this missed by the original query?

Rank according to Cluster Searching

US4393986 – Surfboard carrying rack


Used ‘two wheel vehicle” rather than bicycle in abstract, and picture is of motoycycle. However body of patent clearly notes that the invention is intended for bicycles rather than cycles


US4928863 – Bicycle rack for carrying sport boards



Used the term ‘sport boards’ rather than surfboard in the abstract


US3338484 Load support means



No published abstract (1965 patent)


US3827613 – Golf bag bicycle rack


Referred to golf bag rather than surfboard. But the invention could easily be used to carry a surfboard


US4387836 – Golf bag carrier


As above


US3329323 –  Overhead carrier for use with twowheeled vehicle


No published abstract (1965 patent)


FR709781 – Véhicule pouvant servir au transport de bagages et autres applications


In French language, with no English language abstract



I think this table is interesting at it illuminates what Cluster Searching is trying to achieve – namely to challenge your assumptions on what the answers might be. For example in my preliminary search I had made some assumptions on that the invention might be, or what keywords would be relevant. But these results have come back with otjher ideas – and these other ideas might be exactly what I am looking for.

None of this is intended to provide reasons why you should not continue to use your preferred patent searching process – only that instead that Cluster Searching can provide a quick and useful second opinion that can extend your processes and results.




Want to try cluster Searching for yourself? 

Please contact us and we will be happy to provide a demonstration and free trial


Read More

What is a “Super-Patent”? How would you analyse its litigation risk?

What is a super patent?

Back in 2008 Parchomovsky and Wagner published one of my favourite academic articles on patents, “Patent Portfolios‘, found here. In this paper they introduced the concept of a “Super-Patent”:

“..a collection of closely related patents defining a patent portfolio can be said to operate as a “super-patent” in much the same way that the holding of a U.S. patent grants the right to exclude others from the scope of its claims—the holding of a patent portfolio will allow the holder to exclude others from the collective scope of its claims. Where such patents are both (patentably) distinct yet cover coterminous subject matter,  the breadth of the right to exclude conferred by a patent portfolio is essentially the sum of the individual patent rights.

But the scale advantages of patent portfolios are more than merely additive. The broader protection conferred by patent portfolios offers a range of benefits to the holder different in kind as well as size from a simple collection of unrelated individual patents.” 

With the benefits including:

  • Freedom of movement to innnovate in new areas 
  • Encourage other innovator with work with the patent owners
  • Discourages litigation as it is hard to litigate against a whole portfolio of patents
  • Assists capital raising


How you would analyse the licensing potential and litigation risk of a Super-patent?

The recently developed patent searching tool Cluster Searching is ideal for analysing closely related portolios of patents, i.e. super-patents. As an example, consider the analysis just published by Dr Alex Lee of  TechIPm, of Boston. Dr Lee has used Cluster Searching to analyse what we might consider a ‘super-patent’ (portfolio) of patents for self-driving cars owned by Google – and use the results to predict litigation risks in the area.



Dr Lee has published an number of similar articles, including on patents for smart homes, connected cars and NFC payments.

Ambercite has also published a number of similar assessment, including for a portfolio of high tech batteries, which demonstrates the Licensing Potential analysis available in . As can be seen in this examples, this is analysis is straight forward using Cluster Searching, and ideally suited for working with closely related groups of patents .i.e. “Super-patents”.


Want to try this for yourself?

Ambercite offers free and completely confidential two week trials to Cluster Searching – please contact us to arrange a demonstration and trial. Testimonials for Cluster Searching are found here, including:

Can be a fantastic tool for technical/IP in terms of opportunity finding and to establish the landscape; in terms of how active an area is already and who potentially owns patents in that area, including who files good patents in that area. It’s also good for company landscape searches, and indicating how strong competition is in a particular area” – IP manager, global manufacturing company.


Read More

Can an innovative novel patent searching app strengthen your inter partes review (IPR) petitions?

Demand for inter parte reviews IPRs has been three times higher than the USPTO expected, with 3,277 IPR reviews filed in the last 3 years. This was according to a recent blog published by the director of the USPTO. The director also pointed that to dated only 12 % of claims available to be challenged in the contested patents (4,496 out of 38,462) have been invalidated in a written decision. Other claims were either not challenged, cancelled, resolved by settlement or upheld. This perhaps contrasts with the view of the PTAB as a ‘patent death squad’ as pushed by some.

The 12% ‘success rate’ suggests that the IPR process is far from a walkover. The IPR process allows the petitioner to submit either new or known prior art. Known (or previously cited) prior art is easy to find, but in some cases this can involve reviewing a long lists of prior art citations.

Similarly it can be tempting to look at the published final or non-final rejection reports for the patent to see which prior art was used as the basis of 102 (novelty) or 103 (obviousness) rejection. While this is always going to be an important step, it is entirely possible that the examiner has missed or ignored other cited patents that are also relevant, particularly if combined with other art in an 103 objection,

Alternatively the position of the petitioner can be strengthened by finding new prior art needs to be found, but this can be costly and require time and delays to find.

But what we could both, and in a matter of seconds and using a web application as easy to use as Google patent:


  • Quickly rank what can be a long list of known prior art in a manner that puts both a fresh and objective ranking of the known prior art ?
  • Find not previously cited prior art at the same time?

Would this be a new, disruptive and cost effective process for strengthening the filing IPR petitions?

The pursuit of this objective had led Ambercite to develop the process of Cluster Searching achieve both of these objectives, using a web application as easy to use as Google searching. We will demonstrate this using two case studies recent IPRs.


First case study – US7597502, Jet Blast Resistant Vehicle Arresting Blocks, Beds and Methods

US7597502 is a patent filed by Engineered Arresting Systems Corporation. This patent has a priority date of September 2001, and claims a foam unit for slowing down out of control vehicles, which include a deformable ceramic foam and a frangible cover on top of the foam. Materials referred to in the specification as frangible covers include the likes of cement board.

This patent has attracted a couple of IPRs from a company called Runway Safe.

The patent itself has 24 listed prior art citations. Of course a diligent reviewer would look at every one of these known citations, but it could be helpful to know which of these 24 known prior art citations is most likely to be similar – particularly in cases where there are a lot more than 24 prior art citations.

A search query in Cluster Searching can be as simple as entering US7597502 into Cluster Searching, and setting to date filter to only show patents filed before the patent’s priority date of September 2001. This query looks like this.


And the result? The three most similar patents in ranked order are:

  • US5193764, Aircraft arresting system, filed in 1991. This disclosed the use of a phenolic foam, but without a cover
  • US3066896  Method and means for decelerating aircraft on runways, filed in 1959. This includes a disclosure for a flexible, compressible material with a flexible cover.
  • US5902068, Vehicle arresting unit fabrication methods, filed in Feb 1997. This discloses a concrete foam, which can include a hard coat layer of a thin layer of cellular concrete having a higher dry density.

So how does compare with the examiner found? In this case, the examiner cited US4007917 as being the basis for a 103 objection. This disclosed a ceramic foam covered with a plastic film. This is also relevant, but shows the first benefit of cluster searching – a different ranking is produced (and in an objective manner), which may produce new insights on the known prior art.

And what about unknown (not previously cited) prior art? Near the top of our list of unknown prior art is:

  • US3967704 Vehicle decelerating means, filed in 1971. This discloses a bed of crushable rigid foamed material, which can include a protective covering such as paints, lacquers, bitumen and waxes.
  • GB1122297, Improvements in arrester pads for aircraft runways, filed in 1968. This discloses an arrester pad made from Improvements in arrester pads for aircraft runway, made from honeycomb construction ,..of foam material, synthetic plastics or aerated concrete, along with an upper skin,

So as you can see, Cluster Searching was able to list a series of patents that are very similar to the patent being contested, at least as similar as what is previously known.

You can also see in one of these unknown citaiton documents in a non-US document. You may be surprised how often examiners appear to avoid citing prior art outside of their own jurisdiction. In contrast Cluster Searching can be excellent at finding and ranking highly relevant prior art from other jurisdictions. This is an important benefit. 


Second case study – US6291966, Method and an apparatus for storing and communicating battery information

US6291966 was filed by Ericsson with a priority date of January 1998, and claims a battery with a separate system to log battery information, suitable for use in a mobile phone.

This patent has attracted an IPR from Apple.

The patent itself has just two listed known citations – and yet Cluster Searching was able to return many ‘unknown’ citations in a simple search



Heading the list of unknown (rior art patents returned by this query are

  • US5939856, Battery and charging system using switchable coding devices, which also discloses a battery information circuit, and was filed by Motorola in May 1997.
  • US5606242, Smart battery algorithm for reporting battery parameters to an external device. This was filed by Duracell in October 1994.
  • US6104967, Fault-tolerant battery system employing intra-battery network architecture. This was filed by 3M in July 1997.

And there are many others in this list along a similar vein.

Compared to the just two known prior art patents, this list of unknown and yet potentially relevant prior art patents could help strengthen IPR petitions for this case.


Benefits of Cluster Searching

These two case studies start to illustrate the benefits of cluster searching.

  • An objective perspective is used to rank the known prior art in a new way, so providing new insight on existing information.
  • A list of unknown prior art is also produced. In some cases this can include patents from other jurisdicitons that can also be highly relevant. 

Together, this can greatly expand the patent list for an attorney to base an IPR on. Best of all, this list can be produced in seconds from any computer, and without having to instruct others and wait for the results. Even if other search processes such as outsourcing are used, Cluster Searching can be used to augment relevant patents found by other searchers or processes.


Want to try Cluster Searching 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. You may be surprised (and impressed) by just how much time and money you may save.

Matter pricing is available to patent litigators. 


Read More