Discussion of all things patent mapping and analytics.
Family Cluster Searching has just been released, and this includes a number of benefits - including the ability to search for Chinese, Korean and Japanese patents.
We know this is important for clients due to the huge growth of patent filings in these countries. However patents from these countries can be hard to search for because they may be published in other languages apart from English. Even if translated into English, technical terms can end up with unexpected translations.
Family Cluster Searching offers an alternative approach. Being based on patent citations, it can find links between patents that may not be obvious from keywords or class codes alone.
Imagine, for example, that you were looking for similar patents to CN10424757B, filed by Nike for Footwear having sensor system, in this case a pressure sensor.
We can easily write a query to look for similiar patents, in this case the most similar 2000 patents:
This will returns a list in ranked order headlined by US patents - which is not surprising as this patent also includes a US family member. The top five of these patents in ranked order is shown below:
But - maybe you are more interested in patents only filed in China?
That is OK, we can limit the results to those beginning with CN by applying a filter to the list of country codes:
Which will limit the results to those only filed in China - this is only part of a long list of such patents:
Note the last column, which refers to the type of citation. None of these patents are 'known' citations - all are 'unknown' - in other words none would be recognised as a patent citation using a conventional patent search database.
And are they relevant? The second on this list, for example CN1036378805 is filed for Shoes and method for measuring exercise load - and discloses 'An exercise load measuring device is arranged in a shoe pad inside each shoe body'.
While this particular patents has a priority date after the Nike patent, some other of these Chinese patents have earlier priority dates, for example CN201034771, with a priority date of 2007 and which discloses pressure sensors insides shoes - and again which was not cited by the Nike patent.
What about Japanese, South Korean and Taiwanese patents? And other non-English jurisdictions?
A similar filtering process uncovered 37 Japanese, 6 South Korean and 4 Taiwanese patents, that is patents filed within these countries but not as US or PCT equivalents. We also found 12 German and 7 French patents. Not all were relevant - but many were, and also had earlier priority dates. None of these patent had previously been cited.
But could you find these patents with conventional searching systems?
Maybe - maybe not? This simple search uncovered a lot of patents that had not previously been listed as prior art by any of the examiners of the Nike patent.
What if you had started with the US equivalent Nike patent - being US20130213144
You would have ended up with exactly the same result - Family Cluster Searching provides the same result for each family member.
What if the patent concerned has no US or PCT equivalents
Providing published citaiton data is available to us, we can still end up with a list of similar patents - both in these countries and elsewhere. Those will access to Cluster Searching may want to test this with CN104572956A, filed only in China by Bejing Qihoo Technology Co for a geographic search engine patent. This pulled up a number of Chinese only as well as global patents in the list of known and unknown citations.
Would you like to try this for yourself?
Please contact us for a demonstration and free trial
Special note for Chinese users
The above demonstration refers to lookups in Google patent, which is not available in China. However we offer a Chinese language version that does not rely on Google patents - again please contact us for details.
Update 21 November
A reader wrote to me:
I read your blog post on "A new way to find relevant Chinese ... patents." We know that "similarity" of patents can be defined various ways (or in other words, different similarity measures can be used). In the case of Ambercite's product, is similarity based exclusively on forward and backward citations? Or are other factors like classification or keyword/semantic search part of the formula?
I think this is a great question. We define similarity via a citation based metric, that goes backward and forwards three generation. We like patent citation data because each citation has been applied by a human expert regardless of the keywords or classification codes used - and the latter two methods can introduce errors in your results.
And does it work? No search system is perfect, but we believe Cluster Searching is better than conventional methods in terms of relevance of patents found. Despite this, we still encourage our clients to use Cluster Searching alongside more conventional techniques - because our primary concern is that our users find the best possible results - which is why we developed Cluster Searching.
But there is no need to take our word for it - check out our testimonials, or request a free trial.