How to classify your sentences ?


A) Training via the dialogue box Chat Box


Once the intent of your users known and entered in the “Intents” page, you can begin training your xBot by simply conversing or talking with it in the Chat Box dialogue box.


ChatBoxScoring-en.png


The scoring displayed above the xBot’s response enables you to identify which intent the sentence is the closest to.

By knowing in the chatbox the scores of the intents recognized by the AI, you can know then what action or action the xBot will deliver.


1) How the scoring works ?


  • By default, if scoring is < 50%, the xBot will choose the “Default” Intent set in the page “Knowledge and skills

  • If scoring is > 50% then the xBot will choose the Intent with the highest score and its response will be the one associated to this Intent. This Intent and the associated action or response are set in the dialog editor of the page “Knowledge and skills


2) Recommendation:


If the intent with the high score does not match with the expected result, you can classify manually this sentence in the correct intent on the “Pending” page (Phrases awaiting classification) of the portal and retrain your xBot.


B) Sentences Classification


classification.jpg


You have 4 possibilities to classify a sentence:



The sentence will be used by your xBot to learn that this sentence (and similar sentences) matches the same intent / dialogue scenario.


2) Archive the sentence


When you archive the sentence, you tell your xBot to always ignore this phrase in its learning process. This sentence will no longer appear on the “Pending” page (Phrases awaiting classification).


3) Move the sentence in the category « Test set »


The sentence is categorized in a test set that will allow you to compare your xBot reliably before and after each training.


4) Delete the sentence


The sentence is deleted from the classification. This sentence will therefore appear again in the “Pending” page (= Phrase awaiting classification).



Documented/reviewed 25/07/2017 7:12:55 AM