Welcome to our conversational guide to dialogue management. We'll start with the core terminology and progress to more advanced concepts and strategies for controlling conversations flow and scenarios for using each of them. Feel free to read the contents of this guide in order or jump straight to the section that sparks your interest. Here's a list of topics covered in this guide:
- Definition of Dialogue
- Types of Dialogues
- Definition of Dialogue Management
- Dialogue Management Basics
- Definition of Dialogue Context
- Definition of Dialogue Control
- Definition of Dialogue Initiative
- Types of Dialogue Initiatives
- Definition of Dialogue Flow
- Strategies to Implement Dialogue Flows
- Conclusion
What’s a Dialogue?
A dialogue (also known as conversation) is a form of two-way communication focused on fulfilling an intent. It consists of written or spoken messages frequently exchanged between participants during conversation turns.
What Are the Types of Dialogues?
While there's an unlimited number of possible conversations between customers and conversational bots, looking at them from a high-level perspective allows categorising them into one of the following categories:
What Are Question-Answer-Style Conversations?
Question-answer-style conversations typically happen when customers initiate the dialogue and want to find helpful answers to their questions to satisfy their informational, commercial or navigational intent. They carry little to no context from one conversation turn to another.
What Are Multi-Turn Conversations?
Multi-turn conversations are either business-directed or follow a mixed-initiative. Conversational bots store every answer the consumer gives during this type of dialogue in the conversation context, which helps keep track of the progress and aids the decision-making process.
What’s Dialogue Management?
Dialogue management (also known as conversation management) is a process focused on controlling multi-turn conversations between customers and conversational bots, letting each participant take the initiative to accomplish relevant customer- or business-related intents by managing the flow of dialogues and conversation contexts.
What Are the Dialogue Management Basics?
Although core terminology used in dialogue management might sound familiar, it’s worth quickly going through and understanding the basic definitions because they set the foundations for all the details yet to come. Here’s a list of essential concepts worth noting:
What’s an Utterance?
An utterance is a written or spoken phrase or sentence a potential, or existing customer might use to convey their intent. For example, a consumer might say, “I’d like to book an appointment” or “I want to schedule an appointment”.
What’s a Prompt?
A prompt is a sentence that helps users understand what type of information a conversational bot expects from them to proceed to the next step of the conversation and accomplish the intended goal.
What’s an Action?
An action is a task or activity associated with a particular conversation state representing a step towards fulfilling an intent, such as advancing or branching a conversation, updating relevant records in a database or notifying external systems or users about the outcomes. A specific type of action that accomplishes an intent is a fulfilment action.
What’s a Fulfilment Action?
A fulfilment action is a final step which confirms the completion of a particular customer or business intent, notifies relevant systems and users, and removes its state from active intents stored in the conversation context.
What’s Dialogue Context?
Dialogue context provides all the relevant information about the current conversation. It helps put customer intentions and unique circumstances into perspective, enabling reasoning and influencing the flow of conversations. It consists of four main aspects:
Why Is Interaction History an Important Part of a Dialogue Context?
Consumers remember and often refer to the information shared in the past, making relevant details from current and past interactions critical for genuinely conversational experiences. Conversational bots process this information to personalise their responses.
Why Is It Important to Keep Track of the Current Conversation?
All the information shared recently by the potential or existing customer in the current conversation helps conversational bots put customer utterances into perspective, which is essential for deciding the next best steps. Things like the active and other unfulfilled intents, collected slots, and topics provide a common ground for the upcoming interactions.
Why Is It Important to Keep Track of Past Conversations?
Past interactions are an excellent source of information about client preferences, individual needs, or personal tastes. They help conversational bots adapt the flow of conversations to personalise experiences without forcing customers to repeat themselves.
What’s a Request Context?
The request context contains all the information about the conversation channel and the client's device used for communication, which helps identify its capabilities in the multimodal spectrum and adapt the messages to supported modalities.
What’s a Customer Context?
Customer context consists of personal details that equip conversational bots with client-specific information, enabling them to personalise the experience based on their name, age, sex, location, and other relevant factors.
What’s an Environmental Context?
Environmental context includes all the external information relevant to the current interaction, such as active promotion, time of the day, day of the week, and weather forecast. It enables conversational bots to further personalise experiences for customers.
What’s Dialogue Control?
Dialogue control helps conversational bots to make decisions about the next best steps in the spoken or written conversation after considering customer and business intents in the context of the conversation to manage the flow of interactions and produce beneficial outcomes for both consumers and businesses. There are two vital aspects of conversation control:
What’s a Dialogue Initiative?
Various commercial interactions require participants to take the initiative in the conversation at the right moment to accomplish relevant business or customer intents. A dialogue initiative specifies who controls the flow of the conversation. There are three types of dialogue initiatives.
What Are the Types of Dialogue Initiatives?
There are two scenarios where one of the participants controls the flow of dialogue completely, resulting in the following initiative types: customer- and business-directed conversations. However, there's also a way for them to collaborate categorised as a mixed-initiative dialogue.
What’s a Customer-Directed Dialogue?
In a customer-directed dialogue (also known as customer-initiative dialogue), the customer controls the flow of the conversation by asking informational, commercial or navigational questions and expects relevant answers in response or expressing transactional intent, which indicates that a customer wants to accomplish some task.
What Are the Advantages of Customer-Directed Dialogues?
Letting potential or existing customers take the initiative in conversation and direct the flow of dialogue is essential for allowing them to receive answers to specific questions meant to fill their knowledge gaps or accomplish relevant tasks with the help of your business.
What Are the Disadvantages of Customer-Directed Dialogues?
Most conversational bots can hold conversations only within a specialised area, and letting the customer lead the dialogue in any direction can result in scenarios that cannot be predicted in advance and are thus more difficult to handle effectively.
What’s a Business-Directed Dialogue?
A business-directed dialogue (also known as business-initiative dialogue) is the most common control type in commercial systems. The conversational AI is in control and prompts the consumer for input. In a typical situation, the client expresses their intent using an utterance. A conversational bot then matches business intent and asks questions that enable it to fill the slots and progress in a conversation script to fulfil the business intent.
What Are the Advantages of Business-Directed Dialogues?
Business-initiative dialogues aim to accomplish specific business goals known in advance. They follow predefined paths where customers have to fill in the blanks to progress further and thus are easier to manage than their customer-directed counterparts.
What Are the Disadvantages of Business-Directed Dialogues?
Since business-initiative dialogues follow predictable paths in the form of conversation scripts or slot filling to accomplish relevant business goals, they cannot create genuinely conversational experiences for customers on their own.
What’s a Mixed-Initiative Dialogue?
In a mixed-initiative dialogue, both a business represented by the conversational bot and a customer can take the initiative in the conversation to introduce new, fulfil the currently active, switch to another or resume previous intents at any time.
What Are the Advantages of Mixed-Initiative Dialogues?
Mixed-initiative dialogues are fluid and resemble natural conversations, enabling businesses and customers to ask questions, change topics, and request clarification to fulfil relevant business goals or customer intents.
What Are the Disadvantages of Mixed-Initiative Dialogues?
It’s sometimes challenging to create mechanisms for assessing the relevance of the topic change considering the overall purpose of the dialogue, and decide whether to follow the client’s initiative or stick to a predefined flow of conversation.
What’s Dialogue Flow?
Dialogue flow specifies the decision-making process and implementation strategies used by conversational bots to manage multi-turn conversations with customers ensuring that each conversation feels smooth and comfortable.
What Are the Strategies to Implement Dialogue Flows?
The following strategies help conversational bots to effectively control the flow of dialogues allowing them to transfer the initiative, switch between intents and accomplish relevant business goals and customer intents during natural conversations with consumers:
What Are Conversation Scripts?
Conversation scripts are one of the dialogue flow strategies that contain a set of prompts, advancing and branching rules, and actions after the relevant turns of the conversation. All possible states and the transitions between them are typically presented diagrammatically in the form of a graph.
How Do Conversation Scripts Work?
Conversation scripts rely on state machines that process customer answers based on the conversation context through a set of advancing and branching rules and result in another prompt, re-prompt due to an incomplete or invalid answer, or a fulfilment action.
What’s a State Machine?
A state machine is an abstract mathematical concept that can exist in only one of the predefined states at any given moment. It computes the next state based on the current state and external inputs processed through a set of rules.
What Are Advancing and Branching Rules?
Advancing and branching rules describe conditions for evaluating customer responses in the context of the current conversation, allowing them to progress or branch the conversation along a dialogue path designed to achieve one of the business goals.
How Effective Are Conversation Scripts?
Although conversation scripts have advantages that are especially useful in guiding customers along a predefined path where keeping customers focused on the current step is crucial, they aren’t free from limitations.
What Are the Advantages of Conversation Scripts?
Conversational scripts guide customers by asking them close-ended questions along a predefined path and accepting only valid responses, which helps to keep the focus on the specific aspect needed to progress the conversation further.
What Are the Disadvantages of Conversation Scripts?
The flow of the conversation script is predefined, sometimes making the dialogue feel unnatural or mechanical. It doesn’t have the flexibility of slot filling where the customer can fill in the details in any order and combination.
What Are the Applications of Conversation Scripts?
Typical applications of conversation scripts include leading customers towards accomplishing specific business goals, such as in quizzes used to build scorecards and qualify prospects or questionnaires used for customer surveys and feedback collection.
What’s Slot Filling?
Slot filling is a dialogue flow strategy that controls the flow of conversation by focusing on eliciting required slot values from customers to fulfil their transactional intent. The flow of the dialogue is not predefined like in the conversation script but depends on the clients' responses to the prompts.
How Does Slot Filling Work?
Slot filling starts with matching customer intents and filling required slots with the values elicited from the consumer using prompts to fulfil their transactional intent. It relies on keeping track of the current conversation history and not asking redundant questions.
How Effective Is Slot Filling?
Although slot filling has advantages that are especially useful in fulfilling transactional customer intents where conversational bots collect the required information from clients and trigger a relevant fulfilment action afterwards, it isn’t free from limitations.
What Are the Advantages of Slot Filling?
The slot filling approach depends on the consumer's responses to the prompts and collects relevant details in any order, allowing for better flexibility than conversation scripts. For example, when the customer responds to the initial prompt with more than one appropriate value, the conversational bot can fill all the valid slots and ask only for the remaining ones.
What Are the Disadvantages of Slot Filling?
Since slot filling allows customers to provide the required information in any order, it is more difficult to trigger actions before collecting all necessary details, which can sometimes be a downside compared to conversation scripts that are free from this limitation.
What Are the Applications of Slot Filling?
A typical application of slot filling is collecting relevant information and accomplishing customer tasks. A conversational bot prompts consumers for required slots before fulfilling their transactional intent.
What’s Intent Lifecycle Management?
Intent management is a dialogue flow strategy focused on controlling and supervising the lifecycle of the customer- and business-related intents during spoken or written conversations between potential or existing customers and conversational bots, including mechanisms for starting, switching, resuming and fulfilling intents.
How Does Intent Lifecycle Management Work?
Intent management involves managing the lifecycle of customer and business intents during conversations. Achieving this goal relies on the conversational bot's ability to address the following activities effectively:
How Does Starting a New Intent Work?
A prerequisite for starting a new intent is no active intent in the conversation context. The next step involves matching algorithms that recognise customer intents from their utterances. Sometimes, when the confidence score is low, a conversational bot may ask the client for confirmation. Matching business intents relies on external events and business logic.
How Does Switching Intents Work?
Switching intents is possible when there's an active intent, and either a consumer or a conversational bot decides that another one should take priority. When it happens, the bot needs to save the state of the current intent into the conversation context, enabling resuming it later.
How Does Switching Intents by the Consumer Work?
Switching consumer intents relies on the conversational bot’s ability to understand that the customer’s answer doesn’t satisfy the expectations of the currently active intent. However, instead of simply rejecting it, the bot checks if it matches another customer's intent and, if it does, switches to it.
How Does Switching Intents by the Conversational Bot Work?
Switching intents by conversational bots relies on business intelligence algorithms matching business intents. When a bot decides that a business intent should take priority, it takes the initiative and accomplishes relevant actions before resuming another intent.
How Does Resuming an Intent Work?
Resuming an unfulfilled intent relies on a two-step process. A conversational bot must match it and then retrieve its state from the conversation context to enable the seamless continuation of the dialogue without asking the same questions again.
How Does Fulfilling an Intent Work?
Fulfilling intents involves sending responses and accomplishing relevant tasks at the right moments during natural written or spoken conversations between consumers and conversational bots. Fulfilling different types of customer intents and business intents vary slightly, so here's a list that goes into greater detail:
- Fulfilling Informational, Commercial, and Navigational Customer Intents
- Fulfilling Transactional Customer Intent
- Fulfilling Business Intents
How Does Fulfilling Informational, Commercial and Navigational Customer Intents Work?
Since informational, commercial, and navigational customer intents follow question-answer-style conversations, it's possible to fulfil them immediately by responding with relevant answers to customer questions.
How Does Fulfilling Transactional Customer Intent Work?
Fulfilling transactional intent requires filling in the slots during a multi-step conversation before triggering a fulfilment action. It involves prompting consumers for information, sending confirmations, and re-prompts in case the consumer provides an invalid or incomplete answer.
How Does Fulfilling Business Intents Work?
Fulfilling business intents relies on consumers progressing through conversation scripts and conversational bots firing associated actions at every step in the conversation. This gradual decision-making process allows changing the flow of dialogues at any moment based on customer answers and business intelligence to achieve business goals.
How Effective is Intent Lifecycle Management?
Intent lifecycle management has advantages that enable effective implementation of conversations that unfold naturally between intelligent conversation participants who share common knowledge and can autonomously decide when to take the initiative in the conversation to prioritise another intent, but it's not free from downsides.
What Are the Advantages of Intent Lifecycle Management?
Intent lifecycle management enables changing the active intent of the conversation without forgetting other previously discussed threads that haven't been fulfilled and allows resuming them from where they ended without having to repeat the same parts.
What Are the Disadvantages of Intent Lifecycle Management?
Intent lifecycle management comes at the cost of additional technical complexity. The conversational bot must keep track of the current and other unfulfilled intents to allow seamless state retrieval from the conversation context to enable resuming previous intents.
Why Is Intent Lifecycle Management Important?
Customers might change their minds during a conversation, and enabling them to do so is essential for improving the customer experience. It's also vital to allow the conversational bot to switch intents when it decides that a business intent should take priority.
What Are the Applications of Intent Lifecycle Management?
Typical intent lifecycle management applications include implementing situations where consumers change their minds mid-conversation and mixed-initiative dialogues where conversational bot switches from a customer to business intent to satisfy a business goal before resuming a customer intent.
How to Select a Dialogue Flow Implementation Strategy?
Accomplishing transactional customer intents typically relies on slot filling. Achieving business goals involves guiding potential or existing customers along predefined paths where conversation scripts shine, and enabling a seamless transition between intents and initiatives calls for intent lifecycle management.
Conclusion
Dialogue management is crucial to automating conversations between businesses and their customers. It enables conversational bots to control the flow of dialogues based on customer interactions and conversation context to satisfy customer intents and take the initiative at the right moments to steer the conversations and guide customers along a predefined path to achieve business goals.