Back to article
Title
Publ. No.
Applicant
Priority Date
Similarity
PUBLICATION NO.: 
SIMILARITY RANK
This chart shows the patent families of the result list distributed over their priority year. The chart can be interpreted as the lifecycle of the technology defined by your query and displays the patenting activity of that queried technology over the years, past peaks of activity and revealing trends (is the R&D effort dying or increasing? Is the queried technology outdated, e.g. like DVD or optical disk technology?).
The height of the green bars represents the number of patent families in the result list having the respective priority year.
The highest green bars denote the years with the most numerous or more frequent advances in the result list for the queried technology.
The orange curve shows the similarity average of the patent families assigned to a particular priority year. Peaks in the yellow curve signal the priority years of documents which, at least on average, are likely to be most similar to the technology of your query, i.e. the year when the technology advances had most of the features described by the query.
The slices of this pie chart represent the relative sizes of the patent family portfolios of the top 5 applicant names based on the patent families in the octimine result list.
Click on one slice to filter the result list to the slice’s applicant name. This will display only the patent families concerning that applicant´s name. All other charts will be automatically readjusted to reflect this restriction to a single applicant name.
This chart shows the distribution of the patent families of the result list over the 35 broad technological fields (35TF) defined by WIPO document IPC/CE/41/5 (“WIPO IPC-Technology Concordance Table”).
This simplified technology classification aims at aggregating different IPC-codes into just 35 concisely named classes.
The main advantage here is that it aims to group and display technological areas and instead of search codes. Click on a main category like “Chemistry” to display further 35TF subcategories.
See Ulrich Schmoch’s article for more information: https://www.wipo.int/export/sites/www/ipstats/en/statistics/patents/pdf/wipo_ipc_technology.pdf
Available main areas and technical fields are:
The priority year of a patent relates to the earliest filing year from within a family of patent applications (we use the DOCDB patent family from the European Patent Office). When the matter involves just one patent application, the priority year would be the filing year of the single application. Note, a patent application will be published 18 months after filing.
The filter allows you to screen results within specific time ranges.
The publication year of a patent document is the year in which this patent application document is published by the respective Patent Office. Note, a patent application will be published 18 months after filing.
The filter allows you to screen results within specific time ranges.
Similarity reflects statistical and semantic similarity between documents.
The index range is centered between a minimum of 0 and a maximum of 1000.
Independent of the underlying technology patents with a similarity of less than 100 will not be returned.
A similarity value of 1.000 means the text is not only semantically but also lexically identical.
Due to our complex algorithms the value of the index cannot be interpreted in a strictly linear manner (e.g. a patent pair with a value of 400 is not a third more similar than a patent pair with a value of 300).
The priority date of a patent relates to the earliest filing date from within a family of patent applications (we use the DOCDB patent family from the European Patent Office). When the matter involves just one patent application, the priority date would be the filing date of the single application. Note, a patent application will be published 18 months after filing.
Like publications in the scientific literature, not every patent or invention is equally important or carries the same economic potential. Besides providing good technical search results we also want to give our users some analytics indication how economically valuable and impactful the search results are supposed to be.
From research studies like Gambardella et al (2008) we know that the distribution of the economic patent values is highly skewed. That means a large part of the patent families carries only marginal economic value while a few patent families are of extraordinary high value (e.g. like patents protecting blockbuster pharmaceutical drugs).
The Patent Value indicator measures the economic value or impact of the patent family. It is computed using Octimine proprietary algorithms which have been calibrated based on inventor survey data, auctions, licensing and renewal data.
A patent family with Patent Value in the top 1% is considered to have more economic impact than a patent family with Patent Value bottom 75%.
Our models use up to 50 independent input variables. The predictions are technology as well as time specific. Besides standard predictors like the number of citations, references, claims or GDP (Gross Domestic Product) weighted family size we use technological density and semantic similarity measures to further increase the prediction quality.
There are three significant USPs (Unique Selling Propositions) related to our prediction technology.
First, it is based on strong scientific research, especially the ground-breaking work from world class scientist and Director of the Max-Planck-Institute for Innovation and Competition in Munich, Dietmar Harhoff. His research has been validated in many publications in the most renowned, peer-reviewed scientific journals.
Second, we have access to the data of a unique, proprietary, large-scale and worldwide inventor survey on patent valuation. This gives us a benchmarking dataset, i.e. a ground truth, for patent valuations that we can use to train and test our machine learning based valuation models.
Third, combining patent valuation methods with our semantic technology allows us not only to improve the prediction models, but it also allows us to give patent value estimations in the absence of patents’ forward citations. This means we can compute patent valuations already at the time of publication, there is no need to wait for forward citations.
This indicator is based on cutting-edge research, but it is recommended to use it carefully as it is just an approximation of the true values.
Our Legal Risk indicator describes the probability that a patent family, i.e. at least one family member, gets involved in a legal dispute like an opposition, invalidity or litigation case.
Legal Risk is computed using proprietary algorithms and measures the legal risk incurred by a patent or by members of the patent family of being involved in post-grant review, opposition, invalidity proceedings or litigation.
A family with a Legal Risk in the top 1% has a higher probability than a bottom 75% of triggering a legal dispute.
Our models use up to 50 independent input variables. The predictions are technology as well as time specific. Besides standard predictors like the number of citations, references, claims, patent value, GDP (Gross Domestic Product) weighted family size or the applicants dispute history we use technological density and semantic similarity measures to further increase the prediction quality. Three major factors driving the legal risk estimation are the value of patents, the technological density in the technology field as well as the applicant’s dispute history.
There are two significant USPs (Unique Selling Propositions) related to our prediction technology.
First, it is based on strong scientific research, especially the ground-breaking work from world class scientist and Director of the Max-Planck-Institute for Innovation and Competition in Munich, Dietmar Harhoff. His research on opposition, invalidity and litigation has been validated in many publications in the most renowned, peer-reviewed scientific journals.
Second, combining standard bibliographic patent information with our semantic technology allows us not only to improve the prediction models, but it also allows us to give patent value estimations in the absence of patents’ forward citations. This means we can compute patent valuations at the time of publications, there is no need to wait for forward citations.
This indicator is based on cutting-edge research, but it is recommended to use it carefully as it is just an approximation of the true values.
Innovation Speed measures the median duration between filing dates of patent filings within the family and the overall set of patents referenced in search reports or in the patent filing itself.
Innovation Speed can be of interest to analysts who seek to identify patent families protecting inventions at the cutting-edge of technology development. The fewer the years between the filing date of a patent and its references, the greater the innovation speed.
The indicator is scaled in years.
For example, a patent family with Innovation Speed of 3.5 years indicates a higher innovation speed as a patent family with Innovation Speed of 7 years.
This indicator is based on cutting-edge research, but it is recommended to use it carefully as it is just an approximation of the true values.
Citations or so called forward citations to an underlying patent are patent applications and other publications that were published after the underlying patent, and which cite the underlying patent as a reference in examiner search reports or by the applicant or the inventor during the filing process. A patent receives a citation if they are technically similar to one another or even more if they are building on each other.
This indicator describes the number of citations a patent receives from other patents filed later. Similar like in scientific publications citations describe the knowledge that has been published later and is building on the underlying invention or on the patent. In general, the more citations a patent receives the higher is its technological impact and the stronger it influences later inventions. Patents on broad breakthrough technologies are supposed to receive lots of citations.
The number of citations is one of the most stable and strongest positive correlators with the economic value of a patent.
References of an underlying patent are all earlier patents that are cited as prior art in the examiner search report of this underlying patent published during examination or by the inventor or the applicant during the filing process.
This indicator describes the number of references assigned to a patent. Similar like in scientific publications references describe the prior knowledge an invention or a patent is built on.
The number of references correlates mostly positively with the economic value of a patent.
The refine search feature allows you to make your search result more precise. Assume your query document from Springer Professional is very broad and covers many different topics then it is very likely that also your search results will be broader. That means some results will be relevant for you while others will not.
By selecting up to 5 patents from the results list that match your search expectations you can fine tune your search and get more precise search results. Octimine’s AI technology takes your feedback into consideration, learns from it, combines the semantic meaning of patents and the literature query to raise the search result quality to the next level. This approach is a great example for the power of human intelligence combined with machine intelligence leading to an iterative search process.
Get PDF