Like many others I have been following the promise of AI based patent searching for many years - and with mixed feelings.
On one hand there has been the success of Ambercite - a clever algorithm based on patent citations that can quickly find and rank similar patents, providing that you start with a good starting patent or starting patents.
While Ambercite can be a great solution if you have good starting patents, sometimes this is not the case. For this reason a number of companies have developed and offered language-analysis based AI search products. I have tested a number of these - and often come up disappointed. And this was not surprising - many patent documents comprise a series of technical words combined in similar ways, even for quite different inventions.
However, in recent times Patseer have released a new generation AI search product - and it is rather good.
As an example of how the Patseer AI search tool works, we can look at a recent Australian invention, one for which there is no search report published (at the date that I wrote this blog). Australian patent application AU2023201784 discloses A building that rises with flood waters, or as described in the abstract:
A building that rises with flood waters is disclosed comprising a building base for receiving a building thereon. The building base rises when a surrounding water level rises and falls when the surrounding water level subsides. The building includes an anchoring arrangement for holding the building in position on the ground so that it does not move laterally or sideways across the ground when the water level rises, and a lifting and lowering arrangement extending between the anchoring arrangement and the building base. The lifting and lowering arrangement permits the building base to rise and fall whilst retaining the same horizontal position relative to the ground on which it is mounted. The lifting and lowering arrangement may be positioned directly below the building base and is received within the perimeter of the building base and does not project out laterally beyond the building base. The building may further include a building superstructure mounted on the building base.
So a possible response to increased flooding in some Australian towns.
Lets see what the Patseer AI search capability can do with this abstract. To keep it simple, I pasted the above abstract into the AI search box in Patseer:
And if you want to, you can fine tune these results:
But if you wish, you can simply go with the default settings, which work pretty well.
The first two records look like this:
There are no great surprises with the first results - Patseer has captured the patent abstract. But the second result is very relevant. As a number of other results, for example:
And
So it works, as we hope it would.
How do the results compare with conventional searching? There are many queries we would make to cross-check the results, but this would certainly be a good starting point for building such queries.
Even better,. these results could be used to help build such queries. For example, we could consider the list of common CPC codes:
and topics:
That was easy, wasn’t it?
Now…. none of this takes away the role of the searcher in reviewing and analysing the results. And in fact, the skilled searcher can fine tune the AI results
And this improves the results based on the nominated patents:
Will AI searching replace conventional searching?
Personally I think there will always be a role of conventional patent searching, based on building complex queries with keywords etc.
But this does not diminish the value of AI searching as an additional tool to complement and enhance conventional searching - and in the same way that Ambercite can also extend and enhance conventional searching (Note that Ambercite uses citation networks as the basis of its AI algorithm, rather than language analysis. Which is better? That is like asking which is better - a jigsaw or a skillsaw? To which the answer is - it depends on the job to be done - and a competent carpenter will have and use both - in addition to all of the other tools they use).
So no, I don’t think that AI wil replace conventional searching - but I do think it will play an increasing role in searching. So yes - AI is the future of patent searching.
