Text Input / AI

Exact Comprehension

Here you can specify words or expressions of the user that should be intercepted prioritized to react with a certain behavior. The bot then checks each user input for specific criteria. After recognizing one of the saved phrases in a user input, he continues to execute at a conversation block set by you. The table in the interface shows a list of user inputs which the bot reacts the same way. About New input in the upper right you can create a new procedure.

Properties

  • Name: generic term for all inputs for which the bot should show the same behavior (for example "swear words")
  • Input language: Language that is active during input
  • Category: Group to which the entry should belong
  • Add: ​expressions that lead to the recognition of the input according to certain criteria
  • Block to call: Call block with which the bot will continue after successful recognition
Text Recognition

Add sets the criteria to evaluate user input as valid to continue at your chosen conversation block. Criteria and individual expressions that lead to recognition are specified here.

Recognition Criteria

User input can be checked for 3 different criteria:

  • Starts with: User input must begin with one of the expressions specified here
  • Contains: The user input must contain at least one of the expressions defined here
  • Entire input: Full user input must match one of the expressions specified here. If you choose this detection method, the criteria Contains and Starts with will be disabled.

This example has Contains as chosen input:

Input Text Types

!

  • Text: word or phrase itself
  • Entity: word or phrase of a group of words stored in a word [more about entities]
Creating an exact user input
  1. Open the dialog window via New term in the upper right corner of the interface
  2. Give a name as a generic term
  3. Add one or more expressions that a user needs to make in order for the bot to respond.
  4. Select a conversation block that the bot should continue upon successful discovery.

Loose Comprehension

Using loose text comprehension, the bot tries to understand free user input. This can be used, for example, if the user asks whole questions for which no better recognition has been set [see Exact Comprehension].

Similar to exact text comprehension, you define expressions or sentences per recognition rule that the bot should recognize. This checks the bot with its integrated voice intelligence on similarity with the text entered by the user (free) and, on successful detection, jumps to a specific call block.

Characteristics
  • Name of input group: generic term for all inputs for which the bot should show the same behavior
  • User says: Sets the bot checks for similarity
  • Detection Sensitivity: degree of how "loose" the similarity is.
  • Block to call: Destination call block which will be called after successful recognition

Misunderstood Input

Here all user input that the bot did not understand will be listed , i.e. for which there was no predefined behavior in the bot logic at the time of input (e.g. by loose or exact text comprehension).

The buttons Learn lets the bot "learn" the user input. That is, the misunderstood input is then assigned to a category for which the bot can have a behavioral rule.

Entities

Entities are wildcards, similar to variables, but their values ​​are fixed from scratch and thus can not be changed. An entity (or Word group) contains one or more phrases that in turn contain one or more individual words/phrases. The button New Entity in the upper right corner lets you create new entities.

Each word group has a keyword and may contain synonyms.

Example:

Usage

Entities can be specified in exact comprehension so that the bot checks the keywords and their synonyms from the phrases of the terms for match with user input.

Example:

In addition, (directly) preceding user input can be searched for an entity with the action Find term & save value. The bot then looks for keywords and synonyms in the user text and stores them in a variable [more about this action].

results matching ""

    No results matching ""