When keywords fail...a case study of navigating patent data without keywords

October 23 2018 I supplied some search results to a client the other day. It was one of those difficult searches where a general keyword search produced thousands of hits to go through. Some of these were close to what I looking for - while many others used the same keywords, same class codes - but in some ways were the opposite of what I was looking for.

So what to do? In the end I compromised on a imperfect keyword based query, and manually reviewed several hundred patents. This produced a short list of 20 patents that were close to what I was looking for - but not perfect.

I then used these 20 patents as as the basis of an Ambercite Ai search. And in the 10th ranked patent in the list of similar patents listed by Ambercite, hidden inside the description (but not the abstract or the claims) was the specific disclosure I was searching for.

This knockout patent fell right within the list of class codes I was expecting, but was missing a couple of keywords that I had thought were critical to the search. The concepts underpinning the missing keywords were there though - but were described in a way that completed avoided the expected terminology.

Once again, Ambercite worked in its ability to find highly relevant patents missed by keywords searching

No surprises there - it has worked for many people many times before.


This case study again illustrates the weaknesses of keyword searching. Of course keyword searching will always a useful tool, but anyone who relies on keyword searching alone, whether a patent attorney, patent examiner, patent owner or patent analyst, is in my opinion a very brave person indeed.

There are a many good reasons for the limitations of keyword searching. The include the usual suspects:

  • Variability in the way that concepts are described - a good example is that my ‘carton’ may be your ‘box’. Or along the lines of the case study above, a patent attorney might have even described it as ‘an polygonal object for carrying other objects’ (and I am seeing an increasing number of examples of indirect language in patent applications). Although in many cases this is inadvertent - different patent drafters can describe objects in different ways.

  • The imprecision of language. Sometimes, for example in IT, the same combination of technical words, e.g. ‘database’ and ‘data’ and ‘network’ and ‘processor’ can describe a huge range of very different inventions.

And there is a third more subtle reason, namely the difficulties of the ‘not’ search - as we will discuss below.

Navigation concept.gif

Case study - ‘navigating without GPS' - how to run a ‘not’ search

Imagine, for example, you were asked to search for a patent for a navigation system that does not rely on GPS.

How might you frame such a search query?

You might look for known navigation tools that are not reliant on GPS, i.e. a compass maybe. But this relies on an understanding of what you are looking for - which is not always the case.

Or you could try to somehow create a query, maybe along the lines of:

navigat* NOT GPS.

But what happens if a relevant patent claims ‘a methods of navigation without GPS’?

Maybe you could try:

navigat* AND “without GPS”.

But what happens if a patent discusses ‘ a method of navigating where GPS is not available’?. Or maybe a relevant patent does not refer to GPS at all?

OK, there are many more sophisticated queries that we could write, but hopefully the challenges of NOT queries are becoming clearer.


There is an easier way. Running the query “navigating without GPS” within Google Patent returned 686,000 results. The first few patents listed were patents for navigating with GPS - the imperfection of searching with keywords strikes again…

But in 5th position on the front page was US20070001904A1, filed for a System and method navigating indoors and outdoors without GPS, utilizing a network of sensors.

This appears to be quite relevant - and just as well, as none of the rest of the patents in the first few pages of the Google query were relevant.

This can be part one of the search process.

How Ambercite can find many more patents

Armed with this patent number, it is easy to run an Ambercite search for similar patents:

Navigation patent search.gif

This search returned 50 results as requested, including:

  • #1 ranked patent, US7899583B2 System and method of detecting and navigating to empty parking spaces - “We have focused on the task of detecting and navigating even in situations in which Global Positioning Systems (GPS) cannot provide this information”.

  • #3 ranked patent, US9420423B1 RF beacon deployment and method of use - “The method of claim 2, wherein the geographic region is located in an environment where signals of a global positioning system (GPS) are unavailable”.

  • #4 ranked patent, US8712686B2, System and method for locating, tracking, and/or monitoring the status of personnel and/or assets both indoors and outdoors - “Outdoor tracking methods .. by combining tracking estimates such as inertial tracks with magnetic data, compass data, and/or with GPS, if and when available”.

This is only the first three of the relevant patents I found - and there are many more relevant patents in the list.

So, without having to make any assumptions about what non-GPS technologies I am looking for, or producing a range of clumsy queries, for example:

“not” within 3 words of “GPS” ….”without” within 2 words of “Global Positioning Systems”

Ambercite was quickly able to find highly relevant patents.


Try Ambercite for yourself

Ambercite is available for a free trial (some restrictions apply) where you can see for yourself how easy it can be run similar searches. We can also offer full trials upon request, for qualified candidates.

Learn for yourself why smart patent searchers the world over use Ambercite to enhance their patent searching.

 
Mike Lloyd