To classify articles follow the steps below
-
Create an excel file with the columns
- artId
- query
- conditionId0
- conditionId1
- conditionId2
- conditionId3
- intentName
- isNewIntent (true if you will create a new intent for this query)
- First step is to detect the intention of the query, which means if it fits to an alredy defined intent or a new intent should be created. (if a new intent is created set the field in excel file to true.)
- Align classification with JuliaD
- Go to api.ai (see currentIntents.md for details) --> intent tab on the left panel --> find the name of the intent if you classified the article in an existing one and add manually the query in the Training Phases section.
-
If the query belongs to a new created intent first create the intent (hit + on intent tab in the left side of the panel) and then add the query in the Training Phases section.
- In the query we need to identify the medical entity ( ex. diabetes, stroke etc), and then set this to a parameter sys.any. If there is only one medical condition, the fields condition1, condition2, condition3 can be set to 0 in the excel file.
- Click save on the top right button.
- Make sure to upload the excel file in cloud.zana.com after you classify an article.
- Later this articles will be added to database and intent index.
Important
There are cases when an article needs to be linked with 2 or more medical conditions (usually noticed when 2 or more medical conditions are in a query for example: "What are the causes of stroke and diabetes". In this case we have 2 medical conditions stroke and diabetes.
-
Discuss with JuliaD the condition Ids
- Then the same steps as above should be followed, but in step 5, add parameter sys.any for the first condition, then sys.any1 for the second and so on (example: see Intent searchdifference_condition-comparison).