BOTfriends Preview

Trainings Phrases / Utterances

--> to the BOTwiki - The Chatbot Wiki

Training phrases are a main component of intents. They serve to recognise the previously defined user intentions in the best possible way. For each intent, different question variations are stored in a knowledge base. [1][2][3]

Examples for Training Phrases / Utterances

The following examples show different formulations, all aimed at the same purpose, the weather:

"How will the weather be tomorrow",

"Do I need an umbrella tomorrow?"

"Is there bathing weather tomorrow in Würzburg?"

Scope of Training Phrases / Utterances

How many training phrases are recommended per intent depends strongly on the technology used. The following table shows BOTfriends' experience with the Dialogflow tool.

Number of training phrases/recognition by Dialogflow Intent recognition
< 5 Bad
< 15 alright
< 50 good
< 100 Very good

*It is noted that with Dialogflow more utterances lead to better intent recognition. With other providers, it was found that the quality of the intent recognition decreases again with too many utterances.

Of course, the number of question variations also depends strongly on the use case. In principle, it should be ensured that as many different question variations as possible are included and that the utterances do not overlap with other intents in order to avoid an intent correlation.

> Back to BOTwiki - The Chatbot Wiki

Sources

[1] https://dialogflow.com/docs/intents

[2] https://docs.microsoft.com/en-us/azure/cognitive-services/luis/luis-concept-utterance

[3] https://cloud.ibm.com/docs/services/assistant?topic=assistant-intents


BOTfriends Preview

Human Handover

--> to the BOTwiki - The Chatbot Wiki

A human handover (also human takeover, human handoff) is the forwarding of a conversation from a human to a real person. Chatbot to a real human being.

The term Human Takeover is usually used when the conversation is not handed over, but the person actively takes over a conversation.

Trigger for the Human Handover

A human handover can be triggered by different scenarios:

  • Explicit question of the beneficiaries for a person
  • The chatbot doesn't know the answer to a certain question (default fallback intent is hit)
  • The chatbot is not confident enough (low confidence level)
  • The sentiment of the users shows a negative value (Sentiment Score)
  • A specific intent is made where human intervention is desired or required
  • Certain metrics, such as the shopping cart of an online shop, contain products worth > 1,000 €.

Warm/ Cold Human Handover

A warm handover refers to the immediate forwarding of the user to a staff member. The human response is played out to the user promptly and in the same channel.

A cold handover, on the other hand, interrupts the flow of conversation and/or changes the channel. A common example of this is a handover from Facebook Messenger to the email channel.

Tools for Human Handover

A handover can be integrated into various tools:

> Back to BOTwiki - The Chatbot Wiki


BOTfriends Preview

Tonality of Chatbots

--> to the BOTwiki - The Chatbot Wiki

Chatbot tonality is the tone of voice of the chat bot. For both chatbots and voicebots, tonality determines the style of communication with a user. The tone of voice is a very important part of the Conversational Designs and also plays an important rolein theuser experience ofchatbots.

Relevance of Chatbot Tonality

The chatbot is usually part of an overall communication strategy. For this reason, it is very important to present a consistent image. Since chatbots are conversational user interfaces that are built in dialogue form, some other components have to be considered in comparison to graphical user interfaces (apps, websites). Chatbots are often equipped with characters that are supposed to represent the company or organisation. In addition to the visual component, conversational UIs should also have a particular tone and mode of expression.

Indications of good tonality

  • How does communication with customers currently take place on other channels?
  • How are my company's customers addressed? (You/they)
  • Which language suits my chatbot Use Case?
  • Which target group do I want to address with the chatbot?
  • What values, beliefs and ethics does my company embody in terms of communication?
  • Which guidelines and guidelines already exist with regard to external presentation?
  • Should the chatbot use emojis?

Ultimately, it is important for every company to find an individual way of communicating in other channels in the future. On the one hand the current communication can be maintained and supplemented naturally or on the other hand naturally also completely new ways can be gone and thus its enterprise or its mark more kommunikativer and more innovative to set up.

> Back to BOTwiki - The Chatbot Wiki


BOTfriends Preview

(Default) Fallback Intent

--> to the BOTwiki - The Chatbot Wiki

The fallback intent (also default fallback intent, fallback message, fallback message, error message, fallback interaction) is played if the chatbot cannot assign the user's request to an existing intent or if the confidence score of the assigned intents is too low.

Recommended structure of Fallback Intents

As the Fallback Intent is potentially one of the most played messages, special attention should be paid to the structure and the Conversational Copywriting of the Fallback message:

  • Clearing up the misunderstanding
  • Remind the user of the capabilities and limitations of the chatbot.
  • Call to action: What can the user do next (make suggestions)?

It is equally important to vary the fallback message so that the user does not receive the same message repeatedly.

> Back to BOTwiki - The Chatbot Wiki


BOTfriends Preview

Events (Dialogflow)

--> to the BOTwiki - The Chatbot Wiki

Normally, intents are triggered by user input in the form of natural language. By using events, intents can also be triggered from the backend. In Dialogflow, events can be divided into platform-dependent events and custom events.

An example for events

For a news & media chatbot with push message function, an intent can be stored with an event that sends new articles daily at 6:00 pm. The intent is not triggered after a message from the user, but when the backend system displays a new article on the subscribed topic.

Platform dependent events

This type of event is triggered by user activity on the output channel associated with Dialogflow. A common example is the Welcome Event, which is activated on Facebook after clicking the Go's button.

Custom Events

Custom events are defined for a specific use case and are not dependent on the output channel.

> Back to BOTwiki - The Chatbot Wiki


BOTfriends Preview

Voice Bot / Assistant

--> to the BOTwiki - The Chatbot Wiki

A voice bot is a special form of a conversational user interface and is the counterpart to a chatbot. Conversational user interfaces make it possible to communicate with computer systems in natural language. The input and output of a voice bot is realised via spoken language.

The technology of Voice Bots

The computer is able to convert the incoming speech into text using a Speech-To-Text Converter. The converted text is then interpreted and processed by the system using Natural Language Processing. The output of the speech is done by a Speech-To-Text technology.

For example, the Cloud Services Speech-To-Text and Text-To-Speech from Google and other providers can be used to convert spoken language into text.[1]

Application areas of Voice Bots

Classic voice bots are the Virtual Assistants Alexa from Amazon, Siri from Apple and the Google Assistant, which are mainly operated with voice. However, these can also be addressed by text input. Voice bots can also be found in the smart home area, where they can be used to control the lamps or the heating, for example, using voice commands. [2]

> Back to BOTwiki - The Chatbot Wiki

Sources

[1] https://cloud.google.com/speech-to-text/
[2] Gartner IT Glossary, 2019, "Virtual Assistant".


BOTfriends Preview

Conversational Map / Conversational Flow

--> to the BOTwiki - The Chatbot Wiki

A conversational map is a visual representation of a chatbot conversation. The conversation threads it contains show how the user can go through the individual stages of communication. The map serves as a support in the conception phase and the development by depicting the individual steps of the chat process.

Why is a Conversational Map helpful?

Through the conceptualisation of a chatbot, all stakeholders involved in the project receive an overview of all possible chatbot characteristics and scenarios. On the one hand, this is good for checking whether any content components have been forgotten, and on the other hand, weak points in the conversation structure can be identified very quickly, showing where the user might not get anywhere. With the various communication processes, one speaks of either Happy Paths or Edge Cases. Furthermore, a conversational map helps to play through a classic user journey of a user and to make adjustments if necessary.

Structure of a Conversational Map 

The structure of a project can very well be divided into stages, which already give an overview of the complexity and granularity. There are important components that need to be considered across stages. For example: Where do I use which media (images, buttons, cards, etc.) or which features belong to this conversation step (e.g. sending an email)? Furthermore, you have to think about the style of the chatbot at the beginning. Should the chatbot be a pure click bot or should it be possible to have a free text conversation or a mix of both?

In all cases it is important to think about the answers of the chatbot in each step and to include them in the Conversational Map.

Important levels of a Conversational Map 

  • Welcome Message / Start Message (greeting, address, avatar)
  • Onboarding (clarification of the functionality and the expected content)
  • Content levels (depending on the depth and granularity of the content)
  • Error Message (How does the chatbot react if something went wrong?)
  • Back Message (How can the user navigate back?)

Conversational Map

> Back to BOTwiki - The Chatbot Wiki


BOTfriends Preview

Intents

--> to the BOTwiki - The Chatbot Wiki

Chatbot technologies use machine learning to map the users' natural language to a predefined intention .

Basic principle of Intents

Common systems such as Google Dialogflow [1], IBM Watson Assistant [2] and Microsoft LUIS [3] usually work according to the Intents principle.

Relevant user intentions/intents are defined in advance for the use case of the chatbot. This is done by storing possible user inputs and a corresponding response. User queries are assigned to one or more of the already defined intents on the basis of the user's statement(intent matching). The NLP services usually specify a Confidence Level which indicates how confident the system is in assigning the user's statement to an intent.

Components of an Intent

An intent mainly consists of the following components:

  1. Intent name
  2. utterances
  3. Response

Further components may be added depending on the technology used:

  1. context
  2. Events
  3. actions
  4. Parameters / contained entities
  5. fulfillment

Example Intent

Intent:

Telephone number

Utterances:

"What's your phone number?"

"How can I reach you by phone?"

"Give me the phone number, please."

"I'm afraid I couldn't find your number anywhere. Could you please tell me?"

Response:

"Sure! Our phone number is 0931 123456789."

Important Intents

  • Default Fallback Intent
  • Welcome Intent

> Back to BOTwiki - The Chatbot Wiki

Sources

[1] https://dialogflow.com/docs/intents

[2] https://docs.microsoft.com/en-us/azure/cognitive-services/luis/luis-concept-utterance

[3] https://cloud.ibm.com/docs/services/assistant?topic=assistant-intents


BOTfriends Preview

Entity

--> to the BOTwiki - The Chatbot Wiki

Entities are used to extract user information from natural language.

A distinction is usually made between System Entities and custom entities. System entities are entities already contained in the system for addresses, times and numbers, for example.

Custom entities, on the other hand, can be defined by the user and contain, for example, product information or a staff directory, depending on the use case.

Entities in NLP Services

The Entity Recognition is already integrated in the current NLP Services [1][2][3] and is, besides the Intent Matching, the main component of the Natural Language Processing for chatbots.

Example for an Entity Extraction

I would like to order a small Pizza Margherita to Berliner Straße 1.

Entities:

small (Custom Entity pizza_size)

Margherita (Custom Entity pizza_type)

Berliner Straße 1 (System Entity street_address)

 

> Back to BOTwiki - The Chatbot Wiki

Sources

[1] https://dialogflow.com/docs/intents

[2] https://docs.microsoft.com/en-us/azure/cognitive-services/luis/luis-concept-utterance

[3] https://cloud.ibm.com/docs/services/assistant?topic=assistant-intents


BOTfriends Preview

Confidence Score / Confidence Level

--> to the BOTwiki - The Chatbot Wiki

The Confidence Score, or Classification Threshhold, indicates how sure the NLP service / machine learning model is that the respective intent was correctly assigned. The score can have a value between 0 and 1 , depending on how the neural networks work. In general, a score for each intent is calculated for each user input and the one with the highest value is returned as the result. If the confidence level falls below a predefined limit, a fallback intent is issued.

An example for the calculation of the Confidence Score at Google Dialogflow:

The following four training phrases were entered for the intent "Burger_order". Training phrases have been added:

"Burger Order",

"Order burger",

"I would like to place an order for a burger",

"I want to order a burger"

Dialogflow calculated a confidence score of 0.8 for the user input of "I would like to order a burger from you" . The NLP service Dialogflow is therefore 80% sure that the response issued from the intent "Order_a_burger" was correct. Based on this data and previously defined rules, no fallback intent is issued.

> Back to BOTwiki - The Chatbot Wiki