Classifier score charts available in Cipher and how to interpret them
Classifier Scores Explained
The classifier score is a Cipher derived metric given to patent families which expresses the classifiers confidence that the family is relevant. The score is between 0-1 with 0 suggesting no confidence and 1 suggesting total confidence. Typically the cut off for inclusions in classifier results is 0.5 but this can be adjusted where necessary (or split up as per the below datasets)
Cipher’s Classifier Score Datasets
The ‘classifier score’ section has three datasets allowing different types of analysis. Simply click the classifier score section (below) of your report dashboard to view all of the score related charts.
The datasets are only present if you have run a classifier report (either a landscape or ‘new report’ using one or more classifiers). In a landscape you will only see those of 0.5 or above. In a new report, you are able to see patents between 0-0.5 if you choose to include ‘unrelated’ patents (which fall outside the classifiers)
Families, by score
The above dataset shows the volume of patent families in your report, split by the composite classifier score bands. This chart allows you to click into the various score bands to review patent families. This is useful in different ways, for example with broader technology classifiers, you may want start your analysis with the highly confident results. Conversely those in the lower confidence bands are potentially more likely to show new, unchartered aspects of the technology whilst still retaining relevance.
As above, landscapes will show only those >0.5, however other types of reports will show the full range if you choose to include the ‘unrelated’ patents.
Families, by score and organisation
The above dataset shows the volume of patent families in your report, split by the composite classifier score bands and the organisations included in your report. It will default to the top 10 organisations but this can be expanded with the filters if required.
Families, by score and technology
The above dataset shows the volume of patent families in your report, split by the composite classifier score bands and the different classifiers (or technologies) that you included in your report. It will default to the top 10 technologies but this can be expanded with the filters if required.
Need help? Request support and we'll connect you with a Cipher expert.