What is the Evaluate tab for?
The Evaluate tab is helpful in testing your Classifier, prior to creating any reports. Once you have added a number of examples to your Training Set, you can use the Evaluate tab to perform a 'test run' against a dataset to see what the Classifier finds as a positive match to the technology you are looking for.
How do I perform a test run?
- Select the Evaluate tab and click [ADD]
2. In the pop-up screen that appears, you need to tell Cipher where you would like to apply the Classifier / what data to search. You can use patents from a keyword search, upload a list of patents, or simply select a company (example below):
3. You can then either apply the classifier to a sample set of 1000 patents selected at random, or run the classifier against all patents in your selection (the latter will take longer to produce results). Click 'Apply classifier to selection':
4. Your test run is now complete. Do you agree with how the results have been classified? This will provide you with an indication of much additional training your classifier requires.
Reviewing the 'Positive and borderline' results will show examples of patents that the classifier is still unsure about - make sure to add them as 'Positive' or 'Negative' to assist with the machine's understanding of what you are looking for.
- Do I agree with all the Positive examples found by the classifiers?
- If yes - great, time to create a report!
- If no - review the results and continue adding examples to your training set.
- Rebuild your classifier, and perform another test run.