How novel are smart door locks? A case study of the new Ambercite ‘cluster’ patent searching

最終更新日

Summary

Smart door locks controlled by smartphones may be the next big thing – but are they the next new thing? We explore this question using our newly developed technique of cluster searching, which was able to automatically identify the most similar patents to a shortlist of patents for smart door locks found in a preliminary search – and so augmenting and strengthening this search set. 

And we conclude the blog with highly positive feedback from a patent examiner for a major patent office who was able to use cluster searching to find relevant patents that could not be found using conventional processes. 

Introduction

Smart door locks, controlled from your smartphone, may be the next big thing. As one example the ‘August’ door lock (now available in US Apple stores) replaces the front part of a conventional barrel lock with a device that can be controlled from your smartphone – or you can send e-keys to your friend, for example during a party.

August-lock.jpg

 

Being patent analyts, we were naturally curious to what patent protection the company had. A patent search on Google patent for patents filed by August Home Inc uncovered just two patents, being US2014265359 and WO2014151692.

But is the invention novel? At the date of the publication of this blog, no citations had been published for either of these patents.

This is a very common situation for new inventions, or for patents where citations have been published but opponents wish to invalidate the patent and so may need new patent citations. The typical next step is a patent search based on a combination of keywords and/or patent classes, typically involving many hours of reading the patents found – and a fear that despite that of the work searching these patents, that some key patents had been missed.

And this is a very real fear, as all too often there can highly relevant patents that are missed by the most careful of searchers due to inconsistencies in either keywords or class codes.

Here at Ambercite we have long been aware of the power of citation analysis to find similar patents to an identified patent of interest. We have demonstrated this capability to hundreds of patent searchers, and one of the questions that comes back to us is – what if we have more than one patent of interest?

This is a great question, and reflects how patent searching is carried out in the real world, namely that the searcher narrows done a long list of potential patents to a shortlist of the most relevant patents. Often the searcher will have rated these remaining patents by a relevancy score.

As an example of this, yesterday I carried out an impromptu keyword search using a combination of Google patent and AmberScope for patents that disclosed smart locks that could be controlled by smart phones. This was not a comprehensive search (I will leave this to the experts) but was able to come with the following list of patents

  • WO2012151290, Systems and methods for controlling a locking mechanism using a portable electronic device
  • US7624280, Wireless lock system
  • US8565725, Secure control system for opening locking devices by encrypted acoustic accreditations
  • US8787902, Method for mobile-key service
  • US7114178, Security system
  • US8723641, Access control system and method for operating said system
  • US7196610, Access control system, access control method and devices suitable therefor
  • US8635462, Method and device for managing access control
  • WO2006082526, Method and system for controlling networked wireless locks

Time is short for all of us, and we thought it would be useful if we could use the information that we have already acquired to suggest further patents, i.e. ‘more like this’. Being patent citation analysts, we naturally wanted to do this using citation based methods. For this reason, we have recently developed ‘cluster searching’.

 

Cluster searching

The principle of cluster searching is this. Given a cluster of patents, cluster searching uses citation analysis to suggest other patents that are similar to the the ‘seed’ patents already identified. This is shown conceptually in the figure below, which shows the relationships (via citations connections represented as arrows) between 5 nominated seed patents and another patent, which we will call Found Patent A.

Cluster-searching_20150112-220539_1.jpg

 

In this example, we have shown:

  • A search based on 5 seed patents 
  • That has identifed a ‘Found Patent 1’ that is connected by citation links to 4 of these patents (found patents do not have be connected to all seed patents)
  • And where the line thickness, which represents similarity, is different for the different seed patents.
  • The citation relationships have a direction, but the direction of this citations do not directly feature in this analysis

To identify the relative similarity of found patent A to the cluster of seed patents, we could simply add the similarity values between Found Patent A and the five seed patents. There will also be other other found patents, many other found patents for most cluster searches, and we could use this ‘total similarity’ value to rank these different found patents.

And in fact we could go further. Citation data can be imperfect, which is why we have long discussed the concept of ghost patents, which are similar patents to a given seed patent but which are not directly connected. By searching for ghost patents, we can uncover relevant patents which have been missed or ignored by patent searchers or examiners. We can also include ghost patents in this analysis, to further grow and strengthen our data set. This is shown in the figure below, where Ghost Patent X is an example of a patent which is similar to Found patent 1 and so included in this analysis.

ghost-patent-analysis.jpg

 

And in one further improvement, we could decide that some seed patents are more relevant to the search objective than others, and provide this value with a % scale. Hence seed patents can have relavancy values of 100% if they are completely relevant, or any value from 100% down if they are thought to be less than completely relevant.

But enough of the concepts. Does this all work?

To answer this question, we have applied cluster search to the 8 patents we identified above. We thought that all of the above listed seed patents were pretty relevant, apart from US8565725, which we discounted to a 50% relevancy (as it used sound to transmit locking information, which we think it is a technology detour compared to the more likely transmission of say bluetooth or Near Field Communication).

And what did we find? The 10 most similar found patents are shown in the table below:

Similarity rank

Patent

Title

Similarity score (to seed patents, 100 for most similar patent)

# of citation links to seed patents

AmberScore of found patents

1

US7012503

(2002)

Electronic key device a system and a method of managing electronic key information

100

6

10.7

2

US8482378

(2007)

Access control system and method for operating said system

76

5

0.6

3

US6072402

(1992)

Secure entry system with radio communications

58

5

7.6

4

US8797138

(2009)

One-time access for electronic locking devices

51

4

2.6

5

US4727368

(1986)

Electronic real estate lockbox system

45

5

9.3

6

WO2006082526 (2006)

Method and system for controlling networked wireless locks

43

4

0.4

7

US7061367

(2003)

Managing access to physical assets

42

4

5.7

8

WO200140605

(2000)

An electronic key device, a system and a method of managing electronic key information

30

4

1.9

9

US7457418

(2002)

Method for accessing a user operable device of controlled access

28

3

0.6

10

US5475375

(1993)

Electronic access control systems

26

3

10.6

 

A quick review of these patents shows that all patents relate to using mobile phones to control door locks, which was the subject of the original patent search – and yet describe this concept using a variety of technical terms. 

These top 10 found patents share citation links to at least 3 of the identified seed patents. Of course the full list of related patents is much longer than this, but this should be enough to give you a taste of the results.

We also list AmberScore data in this table, which is an alternative means (if desired) of ranking these results. A higher AmberScore value suggests a more important patent, suggesting that (by coincidence) that the #1 ranked US7012530 has the greatest impact on the field, followed by the #10 patent US5475375.

 

A patent examiner’s viewpoint on cluster searching

We asked a patent examiner working for a major patent office to test our cluster searching algorthm for a patent they were examining. After reviewing back the closest patents to a cluster of patents they had nominated, the examiner noted (I have substituted one of the keywords to preserve confidentiality):

[Top ranked patent] was very good because it found an element that is hard to search for, i.e. ‘independent power settings’.  The ‘independent power settings’ are hard to search for because searching for something like “power near2 settings” would return several thousand hits and most of them would not be for independent power settings and limiting that search with an adjective like “independent” would filter out a lot of good results.  And it’s amazing that it came up first in your search…

Also the top three most relevant patents are owned by [X], and they are all more basic than application I’m looking at, so the chances are high that the company using the application I’m working on owes [X] some licensing fees.

 

Applications of cluster searching

Cluster searching should of particular interest to: 

  • Patent examiners looking to quickly identify similar patents to the most similar patents they have already found
  • Opponents wishing to invalidate a patent
  • Patent owners wishing to license a portfolio of closely related patents – these patents can be the seed patents for the patent search. 

While other patent searching techniques are available, few of these have the potential to provide what can be very similar patents to a list of seed patents, and in a rapid and objective fashion.

Cluster searching is currently being offered as a service, and has already been commercially applied for patent oppositions and similar searches. We are also developing an online version for this search, which should give a real time listing of the most similar patents to seed patents you have found. At the same we will release an API version, which is already attracting interest among existing developers of patent searching databases looking to augment their these databases.

Read More