Workfiles - Sort out relevant documents with an AI-Classifier

Modified on Tue, 5 Nov at 8:30 AM

You can now use AI-Classifier, our new functionality, allowing the selection of relevant documents to reduce the reading time of the experts.


This tool should be activated in a workfile before adding new documents, to automatically classify them with 2 categories : relevant or not relevant.  

You can activate the AI-Classifier on an existing workfile, however it has to be configured and "educated" on a sample before its activation.


As any other AI tools, a learning phase is highly recommended, here in two steps: designate relevant/non relevant documents, and after verify and improve by hand the first run of the tool.

 

How to activate the AI-Classifier on a new workfile:

  1. Create a workfile and activate it by clicking on "Enable auto-classifier" at the top right of the hitlist.
  2. Move/Duplicate at least 10 relevant documents. Note that the tool will learn faster with 50 relevant documents, plus 50 non-relevant documents and if you remain within a particular subject or theme.
  3. Teach the tool what is relevant/non-relevant to you, by selecting, on a few documents, the relevant/non-relevant documents. Use the menu at the top right of the hitlist:
  4. Unselect documents
  5. Then launch the auto-classification from the auto-classifier menu and "run classifier" to sort out the remaining documents 


The procedure to activate the AI-Classifier on an existing workfile starts at the step 3 and remains the same.


Once the process is done on the documents, you will see a score in the Preview tab on the right:

Documents with the "chip" icon  and with a score between 99% and 1% are classified by the Auto-classifier. Documents designated by hand shall have 100% or 0%.


Here you are in the second step of the "tool's education" : you can approve (thus the document will have 100% score) or refuse the classification by switching it to non-relevant (0%).


About the relevancy threshold: this score is calculated by the AI algorithm based on the last run of the auto-classifier. Documents above this score will be seen as "relevant" and below as "non-relevant". 

To change this score, you have to classify/review by hand documents in the workfile. Suggestion: search and review documents around the current threshold. This will have a large impact on the average score and the algorithm will learn faster how to sort relevant documents. 


Once your manual review is done, you have to re-run the auto-classifier on the workfile. The AI algorithm will create a new profile and apply it on the new documents and on the automatically sorted documents.


Once you have taught AI-Classifier to review its choices, it will become more accurate and you will spend less time to review them. Thus sorting out relevant documents for your experts. 


Finally, the Auto-classifier learns differently according to your different working file. A document you consider irrelevant in a specific context but relevant for another can be processed in a different working file.


More tools to help you:

Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select at least one of the reasons
CAPTCHA verification is required.

Feedback sent

We appreciate your effort and will try to fix the article