Infographic
Infographic depicting similar family search result for patent number C0051975087
Identify patent families based on their similarity
Similar family search with respect to patent data refers to a search algorithm or method that is used to identify patent families or related patents based on their similarities. This may involve searching for patents that are similar in terms of their technical specifications, patent claims, or other features.
A patent family includes all of the patents and patent applications that share the same priority application, which means they are related to the same invention. These related patents may be filed in different countries, but they share the same priority date, which is the date of the first patent application filed for that invention.
The purpose of this type of search is to identify other patents that may be relevant to the same technology or invention, which can be useful for patent analysis, and understanding the patent landscape. You can find similar patents from your own unpublished inventions and understand what already exists.
Cipher uses a very sophisticated proprietary patent linguistic algorithm that has been tried and tested over the past 2 years across our Universal Technology Taxonomy (“UTT”) classification. The development of UTT took more than 4 years and is more advanced than most other systems on the market. Therefore, It will typically provide better results than other similarity search tools available on the market.
By leveraging vectorization, we can efficiently calculate the similarity between patent documents and identify similar patents based on their content. This approach allows for fast and scalable processing of large patent datasets.The cost to vectorise all patents is significant, and typically a barrier for suppliers to implement a robust similarity searching system.
Cipher has the advantage of having done much of this vectorisation for UTT, so the incremental workload has been substantially reduced compared to a new entrant developing this.
Where similarity searching is used with multiple source patents, these patents must be related as the algorithm will assume you are providing it with patents about a similar topic. The sophisticated algorithmic approach provides a greater accuracy than similar semantic search driven approach.
For example: imaging looking for “portable solar powered devices.” You can provide one solar powered fan example, and one solar powered pump example, and it will “figure out” that it’s the solar powered bit that’s important (overlap).
Infographic depicting similar family search result for patent number C0051975087
Why has Cipher’s ML suggested these results?
Similarity searching starts with vectorising every patent family in the universe (think of this like
giving each patent family a unique fingerprint).
Each patent family can then have its vector (fingerprint) compared against others to identify vectors
(patent families) that is closest to it, returning the closest results based on the chosen sample size
(50, 100, 1000 etc.).
What parts of the patent are considered when finding a match?
Cipher’s deep learning model (“algorithm”) is specifically designed for patent linguistic tasks and
uses the patent title, abstract, and claims to generate a vector for each individual patent family. It is a
similar process to how the Chat GPT model operates.
What makes Cipher similarity searching better than competing systems?
The processing power, cost, and time required to vectorise all patents are significant, and a barrier for
suppliers to implement a robust similarity searching system. Cipher has leveraged the in-house
vectorisation in classifying every patent family in the world for use in our UTT, so the
testing and heavy lifting have already been done.
Speak to the Cipher team today.