Main Types of Chatbots

First, let me introduce the main types of chatbots.

1) Scenario-type chatbot

A scenario-type chatbot is also called a rule-based chatbot, which responds to questions selected by the user according to predefined scenarios and rules. A scenario with a flowchart structure is created, and the direction of the flowchart branches according to the user's choices. The user can select the appropriate answer from a list of options to get to the necessary information.

As a characteristic of scenario-type chatbots, they cannot answer questions for which they do not have a preference, so they may not be able to reach the final answer. Also, scenario-type chatbots are not aware of "conversations" and use logic to develop them.

It is an advanced version of a FAQ page, so to speak, and is suited for inquiries about handling or setting methods, where the answer to the question has already been determined to some extent.

(2) AI-type chatbots

This is a chatbot service equipped with AI (Artificial Intelligence). The AI picks up keywords, understands the intent of the question, and returns the most appropriate answer.

For example, "What do you recommend today?" What's best?" and so on, it infers needs even from expressions that vary from person to person. During operation, the system can continue to improve its accuracy by learning from accumulated log data, making it applicable to a wider range of situations than the scenario type.

3) Question-and-answer type chatbots

A chatbot that answers a single question with a single answer; if equipped with AI, the AI reads the content of the inquiry and returns a single answer that seems most appropriate; if not equipped with AI, it displays a number of pre-defined questions based on keywords and invites the user to branch out.

There is also a hybrid type of AI and scenario type that guides the user to the appropriate question by branching while answering the question.

Mechanism, features, advantages and disadvantages of each type

The structure, characteristics, merits and demerits of each type are explained in more detail.

1) Scenario Type Chatbots

A database of actual questions and a database listing possible questions and answers are created, and scenarios are designed based on these databases. A tree structure of branching is constructed by matching questions with potential paired answers.

▶Merits

The introduction cost is low compared to AI-type systems. Also, it does not require learning like AI, and can be developed in a short period of time if an FAQ is already available. For industries where the range of inquiries is narrow and limited to certain content, the ability to quickly respond to inquiries can lead to improved customer satisfaction.

Simple inquiries can be resolved by a scenario-based chatbot, and complex inquiries can be passed on to a manned customer support system.

▶Disadvantages

The disadvantage is that the range of inquiries is limited because they are handled within the scope of possible scenarios. If something is not in the configuration, it cannot be answered, and if there is no dictionary, the FAQ tends to become voluminous. Scenarios also tend to be complex, and if the hierarchical structure of the scenario is deep, there is a large risk that users will leave the system before reaching the question.

2) AI-type chatbots

AI picks up keywords from the input conversation to guess and understand the question. Based on learning from data, it statistically selects answers that are most likely to be correct and presents them in text. The system automatically or semi-automatically learns conversation logs while improving the accuracy of correct answers and conversations.

▶ Advantages

More flexible than scenario-based systems for a higher degree of difficulty and a wider range of queries. Combines natural language processing (NLP) and machine learning (ML) techniques, and is designed to allow natural conversations to take place.

Accuracy improves with operation because growth and development can be expected through data accumulation. The system can respond to "fluctuations" in input expressions and correctly read the intent of questions even if there are differences in notation.

▶Disadvantages

The disadvantage is that the introduction cost is higher than that of the scenario type.

3) Question-and-answer type chatbots

It returns a single answer to a single question. As mentioned above, there are two types: AI-equipped type and non-AI-equipped type. However, some services refer to one-answer chatbots as "AI chatbots," so it is necessary to confirm the mechanism before using them.

▶ Advantages

The simple flow of providing a single answer to a single question allows for a user-friendly response. No need to branch out, thus eliminating the need for dialogue scenarios, while at the same time making it easy to set up an exhaustive list of questions.

Users can ask free-form questions without being restricted to a set of options. If the questions are broad and varied, free-entry is more efficient, and the AI can choose to provide the best answer or suggest a list of possible questions.

▶Disadvantages

Depending on whether AI is installed or not, some questions may not be answered depending on how they are phrased. Also, since it cannot have a hierarchical structure, there is a risk of redundant questions and answers for complex questions.

Scenario design for chatbots

This section explains how to design chatbot scenarios and the challenges involved.

How to design a scenario

1. Setting up users

Setting up users determines the situations in which the chatbot will operate and the direction it will take to resolve them. Pursuing a detailed user persona improves ease of use on the user side.

2 . Assuming the question content

Assuming the question content, create a framework for Q&A.

3. Designing branching options

Design the options at the point of branching. At this point, consider the balance between too many branching options and too few options.

4 . Create a scenario for the chatbot

Create a scenario assuming a natural conversation. It is important to keep the exchange straightforward and not too verbose.

5 . Test the scenario

Test it with as many questions as possible and check if you can get smooth answers.

2) Challenges of scenario design

▶ Identifying possible question items and answers

At least 70-80% of the questions and answers should be identified, and the rest should be improved in operation.

▶User demographics and needs

If personas are not fully pursued at the design stage, needs may not be met, leading to user dissatisfaction.

▶ Inadequate or immature scenario design

If scenarios are poorly designed, they will not operate well and the percentage of correct answers to questions will decrease. Also, since the user can only answer according to the scenario that has been created, if the range of assumptions is too narrow, the user may feel unsatisfied.

▶ Lack of flexibility

Because it is difficult to change the scenario once it is decided, it is not easy to respond to changes in the situation or environment.

▶ Difficult to create a perfect scenario

It is difficult to fully automate the process because it is difficult to have all the answers that users require.

Use of chatbots according to needs

Scenario-type chatbots are effective when the scenarios and questions to be handled are very limited, or when they are used as a stopgap measure before transitioning to manned responses for FAQs or customer service. However, if you need to respond to a wide range of questions, AI chatbots are more suitable, although the implementation cost of AI chatbots is higher than scenario chatbots. It also has the advantage of being able to uncover factors that may cause user dissatisfaction and immediately correct them, making it possible to operate with high effectiveness.

The most appropriate chatbot depends on your needs, so it is important to understand the differences between the different types of chatbots. Furthermore, when introducing a chatbot, we should carefully analyze the business needs of your company and choose the method with the highest operational effectiveness.

HUE Chatbot" is a fully-automated AI chatbot that can respond to a wide range of needs, whether it is a question-and-answer type or a scenario type. Please take a look at our introductory materials.