Create a review to mark your chatbot's flow and create new improvement recommendations. If you want to create a review of specific chatbot content, you can create:
- Filtered reviews that exclude specific NLP questions or response passages.
- Targeted reviews of selected passages or conversations.
To create a new review:
- If you want to review:
- All your chatbot content, or all content except specific NLP questions or passages, click Reviews.
- Only passages and conversations you select, click Targeted Reviews.
- Click + Review.
You can also click the empty bar in the Content Coverage or Accuracy charts.
- Select the time period for your review.
- Click Create Review.
It takes a few minutes for all the data to be gathered. The inGenious AI platform will notify you when your review is ready. You can start your review from the Review screen.
Choose your review period
Reviews gather all messages sent from a chatbot user to your chatbot over the specified period. By default, the period starts from the last published version of your chatbot, but you can select any period review, even if it overlaps with a previous review.
If you start a review period before the latest version of your chatbot was published, the review will include passages from the previous version, even if those passages have been deleted in the more recent version. This will not impact your recommendations but may be confusing during the flow marking. The calendar displays when versions of your chatbot were published.
Filter your reviews
You can filter out questions and passages from a review so that you don't have to mark them. This is useful if you want to target your review to a specific section or avoid marking flow for parts of your chatbot that are about to drastically change.
You can filter out chatbot user messages based on the NLP question they triggered, or the passage that the chatbot sent in response. For example:
- You may have an NLP question for common user messages like "Are you a chatbot" that you want to exclude.
- You may want to exclude all the passages where the chatbot greets the user with a wave 👋.
You can add as many filters as you need. Filters are not available in targeted reviews.
To filter out an NLP question from your review:
- Click Filter.
- Select NLP Question.
- Select the question to remove from your review.
All messages from chatbot users that triggered this question will be excluded from the review.
To filter out passages from your review:
- Click Filter.
- Select Regex pattern.
- Type the phrase or regular expression you want to exclude.
The inGenious AI platform will find all passages in your chatbot that match this regular expression. Messages from chatbot users are excluded from your review if the chatbot responded to the message with one of these passages.
Target your reviews
You can target a review to specific conversations or passages you select. This can be helpful for keeping your review sizes manageable, troubleshooting a particular set of questions, and reviewing new or recently updated content.
If you add a passage to a targeted review, the review will include all user messages where the chatbot responded with that passage. If you add a conversation to a targeted review, the review will include all user messages where the chatbot responded with a passage from that conversation. You can add both conversations and passages to the same review and add as many passages and conversations as you need.
To add a conversation to the review:
- Make sure Conversations is selected under Targeted Content.
- Select the conversation from the drop-down.
- To add another conversation, click + Conversation.
All messages from chatbot users that triggered any passage from this conversation will be included in the review.
To add a passage to the review:
- Click Passages under Targeted Content.
- Select the conversation the passage belongs to from the drop-down.
- Select the passage from the drop-down.
- To add another passage, click + Passage.
All messages from chatbot users that triggered this passage will be included in the review.