Tenjin Icon

Flow AI Actions

Query Natural Language

There are two Natural Language processing options available:

  1. Natural Language skill
  2. Query Natural Language action

Most of this document describes the Natural Language skill, the Query Natural Language action is discussed in the final section.

The Natural Language Skill allows you to trigger a project flow based on the text that a user types. Use this skill if you want to trigger different functions of the assistant with natural language.

If the assistant is waiting for the answer to a question, the skill will check if the user’s text matches one of the configured utterances, if it matches, it will call the corresponding project flow. The user’s text does not need to be an exact match to the configured utterance, rather, the skill uses Cognitive Language Understanding (CLU) to determine the match.

If the Natural Language skill is enabled, it is checked anytime a question is answered by the user; with this in mind, please be aware of the following:

  1. If you do not want a particular question’s response to trigger Natural Language, uncheck the Is interruptible toggle in the question’s action.
  2. Natural Language will not trigger a project flow if that same flow is already running; a project flow cannot trigger itself.

The inline help covers the following topics:

  • Examples
  • Testing
  • General Remarks
  • Entities
  • Trying it Out
  • Using Entities in the Project
  • Query Natural Language Action

Text Summarisation

The Text Summarisation flow action shortens long content into a small number of sentences that represent the most important or relevant information from the original content; it works best when supplied with long content, similar to a news article or short document. The output contains the top-ranked sentences within an array, which contains the following:

  • Text: Sentence extracted from the original content.
  • RankScore: The ranked score of the returned sentence, allowing the sentences returned to be ordered by their importance.
  • Offset: Starting position of the sentence within the original content.
  • Length: The number of characters contained within the sentence.

An example of the output from the Text Summarisation action can be found within the inline help.

Single Completion

Use ChatGPT to generate a single completion from a string / one-hit action, such as text summarization, classifying text, analyzing sentiment, or the answer to a simple question.

Assign Chat Completion Message

Assign a messages array, which can later be used in the chat completion flow action, giving the flow developer not only the ability to prepare a new conversation but also to maintain the history of a conversation.

Chat Completion

Once an array of messages has been prepared using the assign chat completion messages flow action, including the system prompt (how the chatbot should behave), any history of the conversation, along with the question being asked by the user, the chat completion action will send the information to ChatGPT and return its response.

Share this article

Comments

0 comments

Article is closed for comments.