How to save money when searching patents – Part 2 – Finding similar and important patents


In part 1 of this series of how use to AmberScope to save money when searching patents, we discussed how to simply review citationally connected patents when searching patent by reviewing each patent.

This is very quick and simple when there are a handful of citationally connected patents. But imagine that there is more than a handful, and in some cases there can be hundreds. Where do you start? How can you make the search exercise simpler, quickly, more effective and cheaper – all at the same time?

The answer is use our series of filters – which this blog will focus on.


Like the previous blog, we will base this search on patent US6705638, which is for an airbag for a glove box. Like before, we can enter this patent into AmberScope (or just click on the image below for an interactive version of the same image).



Filing year filter

At this point in time, it is worthwhile asking – why are search this patent? Typical reasons could include

  • We are trying to invalidate this patent, and are looking for prior art. In this case you would generally be looking for earlier patents
  • We are trying to license this patent. In this case you would generally be looking for later patents
  • We are looking at technology landscape for this general area of technology, and so are interested in similar patents. In the case, the age of the patents may not matter that much.

The reason why ask this is that AmberScope includes a filing year filter to hide patents that fall outside of the age range you are interested in

Lets say, for example, you are only interested in earlier patents. We can adjust teh age filter by moving the orange bars – so that patents filed later than 2002 are hidden:



Which produces a map which looks like this:



Or maybe we are only interested in later patents? So we can set the filing filter to only show patents filed between 2002 and later



So is the filing year filter. For the rest of the blog we are going to assume that you are interested in patents of all ages and so will turn the filong year filter off,  but if not, we can combine the filing year filter with the other filters we are about to describe.


Similarity filter

One of most useful, unique and powerful features in AmberScope is the similarity filter.

To use this, simply one of the following

1) With your mouse, grab the right hand side of the orange circular feature shown here



2) Or enter a new number (lower than 100) into the middle of the circle, say 10. This value of ’10’ tell us that out of all of the patents shown, only the most similar 10% of patents to the US670′ “seed patent” are left in the remaining network.


In this case, you will end up with three direclty connected patents left (coloured blue, purple and orange in the image below, as well as some indirectly connected ghost patents.



And what are these three patents?

  • The blue patent is DE4209604, filed in 1992 by Volkswagen for a Glove compartment cover for motor vehicle – contains airbag and inflator, to restrain occupant during collision,
  • The red patent is US8262126, filed in 2008 by Toyota for an Air bag apparatus for a vehicle (and which refers to an airbag for a glove box)
  • While the orange patent is US8308189, filed in 2009 for a Knee airbag for vehicle (again, an airbag for a glove box)

There are also some indirectly connected but similar patents, which are shown with curved lines from the US670′ patent we started with. Because these slightly translucent, we call these ‘ghost patents’ (which has a double meaning in that they are also hidden from conventional citation analysis).

In this case, the three ghost patent shown are

  • US7261318, filed in 2003 by Autoliv for a Knee airbag and glove box assembly
  • US8596681, filed in 2012 by Nissan for a Internal deployable vehicle panel assembly (a fancy name for a glove box with an airbag)
  • US6971667, filed in 2003 for a Glove compartment airbag system

So, the simple use of this similarity filter quickly found 6 patents all which also disclosed the combination of a glove box and airbag. But not all patents connected in the orignal network to the seed patent show the same. 

This almost eerie ability of the similarity filter to find similar patents to your seed patent is something we have seen again and again. If you have not tested this filter to date, I suggest that you test this when you can – because this filter could you save you a lot of time and money.


Importance filter

How important is a patent? And why this matter?

It is self-evident that all patents are not of equal importance. In our case we have developed the AmberScore metric to apply patent citation information to predict the most important patents. The thinking behind the AmberScore metric is discussed here, but perhaps more importantly is what we can do with the information. In particular, we believe that 

Patents with higher AmberScore values disclose more important inventions, and which can be broader disclosures.

While this rule is not 100% perfect, AmberScore is another useful way of filtering data. The AmberScore for any patent be easily seen in the patent summary box



In this case means that we predict that the patent is 1.7 times as important as the average US patent (AmberScore has been normalised so that the average US patent has an AmberScore value of 1.0.

The importance filter is found to the right of the similarity filter


The bars at either end can be moved to remove the less important (left hand bar) or even the most important (right hand bar) patents. In this case, we might move the bars so that only the most important 10% (or 90% to 100% of the rankings to be precise) of the patents remain


If we do so, only two patents remain besides the focus patent



These patents are: 

  • US7261318, filed in 2003 by Autoliv for a Knee airbag and glove box assembly, which we found in the similarity search. This has an AmberScore value of 5.9, which is quite good.
  • US4552399, filed in 1984 by Honda for a Glove box on vehicular instrument panel, and with an AmberScore of 3.4.

Unlike all of the other patents we have found, the Honda patent does not dislose a combination of a glove box and airbag. The importance filter certainly has its place, particularly if you trying to find the key patents in an area or if you are not sure that you have the right seed patent, but for finding similar patents, not surprisingly we recommend the similar patent filters discussed above.


Keyword filter

AmberScope also includes a simple key word filter, but upfront we need to say that compared to some other keyword filters available to patent searchers this is pretty limited. It applies to all of the data available in a patent summary box,  as shown here:



As you can see, the data available in this summary box is:

  • Patent title
  • Listed patent owner
  • Patent number
  • User comments ( I have the user comment “Relevant to project XX” as an example of this)

So how does our filter work in practice?

Lets say you are interested in which patents are listed as being owned by Toyota? In this case, you might enter the word ‘Toyota’ into the filter box


Found to the right of the other filters



And in this case all of the patents listing “Toyota” into any of the filtered data fields are highlighed with a green circle,



Such as this one here



Note that in the filter box, there were two tabs, with the “Highlight hits” tab selected:



And of course, that is exactly what we have done, highlight the hits for this search with a green circle. The other tab shown in “Hide misses”. If we select this tab instead:



Only the Toyota patents will be shown in the network, along with US670′ seed patent which by design cannot be removed by any of the filters that we have. This is an alternative way of looking for specific patents.



Can we filter on keywords or classification codes?

In contrast to most other patent search engines, AmberScope does not currently offer the more conventional keyword searching in the likes of the the patent abstract, claims or specification.

But why this so? There are three good reasons for this

1) Offering this capability would slow AmberScope down. A lot of people have commented how fast AmberScope is compared to some other patent search engines. There are a number of reasons for this, but one of the key reasons is that we carefully limit the amount of data we send from our server to your computer, since almost nobody has an ideally fast internet connection. This is why we leave out patent abstracts for example – this would slow everything down.

2) A number of other patent search engines do offer this capability

3) We also think that filtering patents by keywords can be very risky and unproductive, and will almost see you both miss relevant patents and find irrelevant. I know everybody else does it, but this is because in most cases they have few other ways of filtering patents (this is a little like how almost everybody used to walk everywhere – until somebody invented the car). We instead think that the AmberScope filters can be a much more effective and productive filtering method.

The same applies to patent classification codes as well.

In fact, a combination of AmberScope and traditional patent searching can work, but this is best done in series and not in parallel, for reasons explained here.


Is this same material available in video?

Yes – this material is repeated at the YouTube video found here.


Read more