Patent Analytics Through Reverse Image Search Engines
Patent Searching – Introduction and Importance
This is the “Century of Technology”. Patents are now seen as an enormous repository of human innovation and knowledge, as well as a class of property that can be owned, traded and leveraged. Intellectual property (IP) equips businesses with an exclusive right, and at the same time prevents illegal usage and/or infringement of their IP. It is a common practice in R&D centers for an innovator to do a preliminary patent search for his invention. However, he may search and find nothing even when there are things that would be found by a professional searcher. The advancement of technology through “incremental inventions” has made it a tough job for innovators, and an even tougher job for professional searchers to identify appropriate prior arts in a fixed time frame.
Companies/Innovators like IBM, Google, Samsung and Apple, are pushing limits for obtaining Intellectual Property under their name. In the US alone, the year 2015 witnessed 629,647 patent applications which makes it more than 1700+ patent applications filed per day . The US patent office granted 325,979 patents in 2015 which makes it close to 900 grants per day.
Armed with complete and accurate information about the universe of patents, business leaders can and should leverage patent analysis to inform a myriad of business decisions – such as acquisitions, competitive monitoring, R&D planning, and strategic decision-making. Apple, Google and Samsung have already set big examples through their recent lawsuits. In the IP industry, the slightest detail could make the difference in million-dollar lawsuits. Hence it is needless to say that a robust, efficient and quick search mechanism for identifying patents can make a huge difference to businesses.
Evolution of Patent Searching – Approaches and Tools
From searching patents manually in the library of patent offices – to the use of complex algorithms; patent searching has indeed matured to a great extent. A new generation of patent analysis tools, applying big-data analytics, cloud technologies, modern software development, and data-improvement techniques have emerged. With the advent of Symantec-Search algorithms developers have marked a new milestone in the patent searching algorithms.
A patent may be broadly classified as a Utility patent or a Design Patent. Accordingly, a searcher may perform a utility patent search or a design patent search to identify relevant prior arts. Since patent documents are classified (IPCs & CPCs), searching for a given concept must not be limited to a keyword search. A concept can perfectly be expressed by a classification symbol. Utility patents are classified into CPCs and National Patent Classes. Design patents are generally classified into design classes like National Design Classes (US, CA, JP) and Locarno Classes. As a best practice, patents should be searched using a combination of keywords and classes. Patent search databases/tools like Questel Orbit, Thomson Innovation, WIPO Global Design Database and so on have these capabilities to combine concepts in a user desirable manner for performing a prior art search. [Explore our Service:
The Next-Gen Patent Searching – A Concept
- Wouldn’t it be nice if the patent analysts/examiners are given a tool wherein they can upload images of the invention to be examined and the tool may retrieve similar patents filed worldwide?
- Wouldn’t it be even better if that tool can accept hand drawn images and return visually similar patents filed worldwide?Sounds good. Right?
An Introduction to Reverse Image Searching
Image search engines (reverse image search engines) are those special kinds of search engines where you don’t need to input any keyword to find pictures. Instead, you need to place a picture and the engine finds the images that are similar to the one you entered.
When you search using an image, your search results may include:
- Similar images
- Sites that include the image
- Other sizes of the image you searched for
Search using an image works best when the image is likely to show up in other places on the web. So, you’ll get more results for famous landmarks than you will for personal images, like your latest family photo.
However, the real task is identifying the correct document based on the image uploaded on the image search tool. You can very well retrieve visually similar images and documents comprising those images but establishing their relevance needs manual analysis.
Reverse Image Search Engines : Usage
Even though the first Reverse Image Search (RIS) tool from Google was first introduced in 2001, RIS tools have not received a global acceptance for patent searching so far. This may be due to the lack of awareness or due to lack of appropriate search algorithms used in these tools.
Above we covered a brief introduction to patent searching and its importance to businesses. We were also introduced to the concept of Reverse Image Search. Let us now take a closer look at the Reverse Image Search tools and their application within the patent search and analytics domain, as well as their limitations/challenges.
Prior-Art Searching – Methodology
As a preliminary step, searchers generally understand a technology based on the images and associated text describing the invention through that image. In case of an identified prior art, the analysts establish relevancy of the patent based on image-based contextual analysis. This approach is generally prevalent on inventions related to mechanical and allied domains. Thus, identifying an image that is visually similar to the invention ensures a better relevancy in the search and reduces effort and time required to identify relevant results. However, manual analysis is still required to establish the relevancy of the results.
Reverse Image Search Engines – Tools
Reverse image search engines are those special kinds of search engines where you have to upload a picture and the engine finds the images similar to one you entered. Thus, you can get to know everything you wish to, simply with the help of one picture.
Search Engines like Google, Bing, and Yandex achieve this by evaluating the submitted picture and creating a calculated model of it using advanced procedures. It is then matched with numerous other photos or images in Google, Bing or Yandex databases before returning, matching and providing you with similar results. It should then be noted that whenever available, Google also makes use of meta-data based on image description.
Google Images were first introduced in 2001 and since then it has been helping millions of users to find the appropriate image results that they are looking for with the help of keywords. More than 90% of us must have used Google Images or any other image search engine, at one point or another, for various purposes.
Other similar Reverse Image Search Engines are:
- Tiltomo (Flickr based tool)
- com (by Ideeinc.com)
- Byo image search (also maintained by Ideeinc.com)
With a myriad of reverse image search engines available globally, it’s a tough choice for inventors, professional searchers, and patent examiners to decide which tool to choose.
Reverse Image Search Engines : Use for Prior-Art Searching
We have used “Google Images” to give you an overview on the search methodology. An example is enclosed here for better understanding. (Please note that the example cited below is exemplary in nature and has no association to any organisation/project).
Fig. 1 – Drag and Drop an image on “Google Images”.
Fig. 2 – Results retrieved through “Google Image” search.
Fig. 3 – Sample patent US 5,040,485
Through the above example, I have tried to explain how RIS engines can be helpful in searching patents by utilizing images. It will be evident to a professional searcher that RIS engines can indeed be very helpful in prior-art search (especially FTO, validity and invalidity searches), where a reference image/patent document is provided for searching similar technologies. In the above example, we successfully identified the patent through its image using RIS engine, however, there still lies a scope of improvement in these engines, vis-a-vis, searching patent literature through product images, hand sketches, cycle diagrams and so forth.
Nowadays, a great deal of emphasis is being laid on automation. There are millions of sectors where “bots” have replaced human involvement to a minimalistic extent. With the help of EDA tools and keyword-based analytics, the field of patent searching & analytics is advancing at a rapid pace. However, searching prior arts using images is still a relatively untouched domain. Even though there is a multitude of tools that host reverse image search engines, their application is still limited. Furthermore, the documents retrieved through reverse image search are generally not as relevant as those retrieved through “concept based” searching. Thus, there stands a huge scope in maturing this technology and applying it to the domain of professional patent searching. Needless to say, involvement of human analytics can never be done away with when it comes down to a prior-art search.
– Tanmay Mittal and the Editorial Team