NLU Activity History

The NLU Activity History feature provides you with insights on how the NLP engine was used in a conversation: how much the NLP engine was used and the NLP scores. Analyzing the NLU Activity History helps you identify the messages with a low NLP score and add the user messages as flow training phrases on the flows where the score was low. This enables you to further fine-tune the bot's ML model.

To access the NLU Activity History, on the main menu, click History & Logs > NLU Activity History. The NLU Activity History page appears which provides you with extensive filtering options.

Selecting a Bot is mandatory to view the NLU activity history. The other filtering options are optional and you can use them as best suits your needs. Select the desired filtering options and at the top-right corner of the page, click the Refresh button. The table displays all user messages matching the filtering criteria with the NLP match scores and the flows where the match is done.

The table below describes the table columns.

Column

Description
Channel The icon in the first column indicates the channel used by the user to chat with the bot. In the figure above, the channel used is web.
User name The name of the user who chatted with the bot.
Time The date and duration of the chat conversation.
Status

Indicates the match type for a user message:

  • ExactMatch - the user message (intent) matches the utterance on a flow. The match is not done via the NLP engine.
  • NoMatch – the NLP engine does not find any matches.
  • SingleMatch – the NLP engine finds a match (partial match) on a single flow.
  • MultipleMatch – the NLP engine finds utterance matches on multiple flows.
Score

The score for matching what the user says with utterances defined on flows.

Based on the Status, the following scores as shown:

  • ExactMatch - score 1.
  • NoMatch – score 0.
  • SingleMatch – NLP match score. The score of the NLP engine for matching the user message with a single flow utterance.

  • MultipleMatch – highest NLP match score. The highest score of the NLP engine for matching the user message with flow utterrance.
Flow1Name The name of the flow with the highest NLP score (if status is MultipleMatch) or if the Status is ExactMatch, the name of the flow which has the utterance matching what the user message. For ExactMatch, the NLP engine does nothing, it's not involved in matching what the user says.
Flow1Score The NLP score if there is
Flow2Name Only if Status is MultipleMatch! If the NLP engine matches the user message with the utterance on two or more than two flows, the name of the flow with the second NLP match score.
Flow2Score Only if Status is MultipleMatch! The NLP match score of the second flow which has the utterance matching the user message.
Flow3Name Only if Status is MultipleMatch! If the NLP engine matches the user message with the utterance on three or more flows, the name of the flow with the third NLP match score.
Flow3Score The NLP match score of the third flow which has the utterance matching the user message.