Our algorithms and data
We use the power of citation analysis to predict patent importance and similarity
It can be hard to judge which patents are more important than others, and a range of measures exists. At Ambercite, we think that patent citations can be vital predictors of patent importance and potential value. Citation relationships signify:
- Activity in an area of importance to one or more applicants; the more activity, the more valuable the area.
- Similarity assigned by a patent examiner or patent owner who felt that the two patents were similar.
AmberScore, a unique measure of patent importance
AmberScore predicts the likely commercial importance of a patent based on the number and strength of its forward, backward and indirect citations. Forward citations, particular recent forward citations are given a higher weighting. Overall AmberScore can help suggest the importance of a patent that takes into account not only the patent itself but also the patent activity in the immediate vicinity of the patent. A higher amount of patent activity is likely to to be caused by a commercially important technology.
AmberScore has been normalised so that the average granted US patent filed in the last 20 years has a value of 1. Higher is better. We have seen some patents with an AmberScore value of 150 or more, but the vast majority of patents range between 0 and 3. AmberScore values can change over time as new patents and citation connections are made.
Why patent ranking metrics such as AmberScore can make some people nervous, we believe that such measures are in fact less risky than treating all patents as having the same commercial value - which in turn can introduce other errors into patent analysis.
Using Amberscore in practice
We provide AmberScore in each of our Automated Reports as well as our search tool, AmberScope. If you use AmberScope, the value is displayed in the details box, as in the image for US7479979. This famous 'Steve Jobs Patent’, has an AmberScore value of 30, or 30 times the average for US patents. AmberScore also governs the size of the node or patent bubble, in our patent search tool..
Tips for benchmarking using AmberScore
We recommend that you consider the jurisdiction, technology and age of a patent when comparing AmberScore values. For this reason it is best to benchmark EP patents to EP patents, or US patents to US patents. Similarly, expert judgement should be applied when comparing the AmberScore value between patents in different areas of technology.
AmberScore values can change over time as more patents and citations are added to the network.
We can find the most similar and relevant patents to your patent of interest – using only citation data to find these patents. In this analysis we disregard keywords or IPC classes as we believe applicants and examiners will have already considered a range of keywords, language differences or IPC classes when making citation judgments. To measure patent pair similarity we consider all surrounding citation judgments, direct and indirect.
Most similar backward or forward citations
Determines which are the most similar known backward or forward citations. In simplified terms this algorithm considers the number of cross-referenced patent documents between patent pairs. Some patents can have hundreds of patent citations, and this algorithm can be useful to prioritize prior art, detect potential infringement or potential licensing opportunities. In our patent search tool AmberScope, thicker lines are used to suggest more similar patents.
'Ghost patents' - similar yet unknown patents
Ambercite can find most the similar patents missed by patent examiners or applicants. The Ghost Patent algorithm was developed to understand, which patents are similar and should have been picked up by the examiner or an applicant. The ghost algorithm assesses first, second and third order citation relationships that surround the seed patent. It considers the the number of cross-references as well as patent importance defined by the Amberscore (see above). We believe in the wealth of human intelligence captured as citation data to give you unique insight into new prior art or licensing opportunities.
Ambercite receives its raw data from one of the world's leading sources of patent data. We then further process and augment the data to meet our standards and requirements.
This meant building a database that comprises:
- 65 million unique patents records. Every patent record is assigned an AmberScore value to reflect is potential importance. Patent applications are grouped with the granted versions of the same patent.
- 135 million citation links, with every link weighted to suggest similarity between the patents they connect.
Our data is updated regularly.
AmberScope and our Automated Reports can work well on patents from many countries, but are particularly effective for patents from the following countries and jurisdictions: