AI Classifier - Moderation Mode

In this mode the records that have been predicted will be provided to you for review, basically the categories that have been assigned you can cross check the categories, modify it or remove it completely and then move further to the next record.

Steps to run the AI classifier in the moderation mode

1. Run a search query and push all / results of choice to a project. 
2. Once a project is created, create categories in your project which are based on the applications of the technology, on the important aspects or the key features of the patent documents etc. 
3. There needs to be a the minimum of 5 records per category, however, we recommend the users to at least keep 10-15 records per category.
4. Now to run the AI Classifier launch the tool either from the Categories filter or from the AI Tools dropdown

5.  Select the Automatic - AI Classifier option

6. Once Automatic Classifier is selected, select the training data set that you have created for the project on which you are running the AI Classifier or  you can use categories that you have created in an other related project.

7. If you have a project in which you have already categorized the records or you have already run the AI classifier, then this dataset can be used as a training model for the project on which you are working currently. The requirement here is that you will have to create similar categories in the project that you are working on currently.

8.  Then in AI Classifier mode select With Moderation and then select the threshold value and click on Confirm

9. Once the AI Classifier is done predicting categories you will see a Blue Column called as Pred Categories as below:

The Pred Categories Columns shows:

View - This dropdown will have 3 options (all these will be shown in the Predictions section besides the category names)

• Predictions -  This option will show you the number of records predicted in each category
• Pending - The records for which you have not accepted the predictions 
• Accepted - The categories for which you have accepted the predictions 

Cutoff: This option will show you the threshold on which your AI Classifier predicted the categories (hover over). Here you also get a MIN and MAX option where you can select a range of threshold and on selection of values in MIN and MAX the records falling in the selected range would be filtered. 

Predictions: This option shows the categories for which the predictions have been done. When a category is selected a filter is applied similar to the one in Cutoff showing the records predicted in the selected category. 

• Once the AI classifier has finished predicting the categories in Moderation mode, the categories will be mentioned below the records in Orange color and a prediction percentage besides the category name for all the Standard Views and Compact View. 

• Now here when you click on dropdown there are 2 options Accept & Accept All. The Accept option will add that record to the said category. However, Accept All will open a box where you can give a threshold value and all the predicted records falling at the selected threshold and above will be added to that category.

• Similar to the Accept All option as described above you will also be able to do that for the Accept Predictions option in Edit dropdown.

Notes:

• For moderation mode the default threshold value shown will be 30
• The cutoff option dropdown in the Min value will show the threshold on which the AI classifier was run 
• While initiating the AI Classifier on your project, if the source project (training model) will not have categories then the confirm option will be disabled.
• If the source project has categories but the destination project does not then the AI Classifier will create categories in the destination project.  
• Creation of categories and categorization of records to the same are a must have for a project on which the AI classifier is to be enabled. 
• Available only in the ProX edition of PatSeer