How can I validate my patent search? When have I searched enough to have confidence in my results?

A recent USPTO tender for the application of artificial intelligence in patent searching asked, among other questions, about how AI can assist with “Capturing search “journey” and identifying when it is complete.”

Personally, I thought that was an excellent question that expressed one of the big questions in patent searching - how do we know when a search is complete? Of course such questions are often bound by search budgets - few searchers have the luxury of an unlimited search budget. But even in cost constrained searches this is a common question.

Take for example, PCT patent application WO2018048951, filed by Fitbit for Methods and systems for labeling sleep states, and with a priority date of September 2016. This claims the identification of sleep states using movement and pulse data (where pulse data is captured using optical sensors).

Fiit sleep states.gif

This patent application has been examined by the WO examining authority, with two listed patent citations, being US2016007934 (=US9808185) and WO2015119726.

I don’t know what search process the examiner used, or how they decided to nominate these prior art patents. But lets assume that it was a rigorous process.

Getting back to the opening question, how might the examiner validate their search process? And so have sufficient confidence that their search is complete?

Ambercite has a unique perspective on patent searching, that can provide a second opinion to your existing search - and so either provide confidence that you have found all of the relevant art, or perhaps suggest new prior art. Ambercite does this by applying a unique AI algorithm to a vast network of patent citations to identify the most similar patents to one or more starting patents - and so exploiting the ‘collective wisdom’ of patents examiners and applicants.

The process of using Ambercite to double-check this patent search is very easy. Simply enter the patents that have been found into Ambercite:

Sleep search query 1.gif

Often in Ambercite searching such a simple query is sufficient, but a quick review of the results suggests that the results are a combination of patents for monitoring sleep, and patents for monitoring other activities, as shown below. Perhaps a slightly more focused search can be justified.

First set of results.gif

So we might expand the patent search from the default most similar 50 up to the most similar 250 patents instead:

Sleep search query 2.gif

And then apply a simple keyword filter of ‘sleep’ in the titles and abstracts of the 250 patent families found:

Sleep filter.gif

This filtered query produced 67 results, which can be reviewed in this interactive HTML file. Among these patents is US98202680, as summarized below.

sleep patent found.gif

I can see the potential relevance of this patent as prior art, but maybe the examiner has already reviewed this patent in more detail, and in fact decided it was not relevant - and therefore did not want to see this again. How could they avoid having to cover this old ground again?

In this case, there is the ability to customise the search so that this patent, and others the examiner has also reviewed and decided to discard, can be left out of the results, namely using the hide box

Hide box.gif

Using this hide box will ensure that only new and potentially relevant patents will be shown - the patents in the hide box (and their family members) will be left out of the listed results - although they will still form part of the network we used to find similar patents.

Incorporating this validation process into your routine searching process.

As you can see, this is a straight-forward process:

1) Identify the best patents you have found in your search

2) Copy these into the Ambercite search engine - and optionally copy the other patents you have found into the hide box

3) Run a patent search, and look for similar patents. This may involve some simply filtering as shown above

4) Review the results. Have you found better prior art?

  • If NO, that is great - your earlier search appears to have been successful.

  • If YES, that is also great - you have just improved your search.

So you can’t lose really - an investment of a just a few minutes can provide the confidence you need to validate your patent search.

Want to try Ambercite for yourself?

Check out the free trials offered here.

Mike Lloyd