사용자의 질문 의도를 이해하여 고객 대응 등 다양한 서비스에 활용할 수 있는 챗봇을 손쉽게 만들 수 있습니다.
Fast and Accurate Chatbot Dialogue Model Linked with Various Messaging Channels
The Chatbot service is equipped with a powerful dialogue model that is built on Naver's data and know-how to help you develop a chatbot that will help satisfy your customer’s needs.
- Smart Chatbot Engine
- The chatbot engine learns like how humans learn to talk by creating dialogue models and learning from them.
- Studies various sentences and grammar
- Remembers important information
- Additionally checks the context before and after
- Additionally performs iterative learning
- Provides feedback based on correct answers.
Then it extracts a reply based on the majority of results using various models.
Based on Naver's know-how, a complicated and delicate process is carried out for the chatbot engine to understand the natural language and handle the replies.
- Multilingual Support (English/Chinese/Japanese)
- The chatbot engine, which is infused with Naver's expertise, is applied with the natural language processing technology and machine learning based algorithm. Based on these technologies,
we are also providing a multilingual chatbot service which supports English, Chinese, and Japanese in addition to Korean.
The learning algorithm is optimized for the characteristics of each language, enabling the use of chatbot services to customers from various countries. And, as a result, the performance of the engine is continuously improved.
- Easy Chatbot Builder
- In order to conduct chatbot learning, "learning data" must be prepared in advance. The learning data is composed of many utterance examples of questions and answers. The chatbot builder allows you to easily manage learning data and test utterances. If a wrong answer is given during the chatbot building test, tune the learning data or make the chatbot re-learn the material in order to teach the chatbot how to find the correct answers. Also, a separate server is not required because the chatbot builder is provided based on cloud.
- Link with Various Channels
- Once a chatbot domain is created, it is possible to link to various channels like messengers or custom channels.
Utterances from various channels are handled real-time so you can easily link to LINE, Tok-Tok, KakaoTalk, and Facebook. You can also link the chatbot easily to web-based pages and mobile applications, such as the client's portals or applications.
In other words, you can easily link to the legacy systems because the chatbot provides specifications that can extend the REST API-based custom endpoint.
- Rich Reply Composites
- NAVER CLOUD PLATFORM provides reply composites so you can flexibly generate chatbot replies. In other words, we not only provide simple text replies but rich reply composites that include text, buttons, multi-button, image, and carousel. You can set up to use these composites for the messenger by clicking a few buttons on the chatbot of messenger. A combination of composites can be created and are conveniently converted to each of the specifications of the linked messenger.
- Functional Dialogue Components
- You can link the legacy system with the chatbot service to reorganize the replies before providing it, create a flow to follow a specific scenario, or model a dialogue to fill up the slot. You can also call the external API via the action method and include it in the reply. Or, you can provide multi-choice or short answer questions. The task is used for executing complicated orders or logic. (You can execute an action by filling up all empty slots, such as ordering pizza or returning goods.)
Naver's natural language processing technology clearly identifies user intents which lead to natural dialogue. The application of a deep self-learning algorithm helps to reduce the need for a high number of user utterance examples and optimizes the dialogue model.
Chatbot Building for Anyone
Even if you are not a software developer, you can easily create a simple chatbot service which handles customer inquiries or can have a basic conversation with a user, and it doesn't even require a lot of time or money. The development tool for chatbot (Chatbot Builder) is provided on cloud so you do not need a separate server nor do you have to develop software.
Effective Dialogue Modeling and Continuous Relearning
In order to make a natural conversation that is not simply a preset scenario, the chatbot must continuously learn the user utterances. NAVER CLOUD PLATFORM's Chatbot collects the utterance patterns of various users to define and learn how to reply to utterances within the same pattern. To do so, it uses input utterance examples with the same intention expressed differently. You can improve the dialogue model by adding utterance examples and making the Chatbot relearn each time a new expression is created or when a problematic expression is discovered.
Abundant Dictionary Data
For a chatbot to correctly identify the user intent in a question, it requires tagging entities and learning on similar patterns from the large amount of utterance examples. NAVER CLOUD PLATFORM's Chatbot provides automatic entity mapping using Naver's abundant dictionary data. It allows for effective tagging since only the entities of a specific domain are extracted and tagged for accuracy.
Users may use the service differently from the utterance patterns predicted and reviewed in the scenario modeling. As modeling cannot provide all possible utterance examples of users, statistics and analysis on input user utterance is very important for an effective service. The NAVER CLOUD PLATFORM provides various types of analysis indicators in order to continue improving the model.
Pre-built templates are provided for customer support or for simple conversations. When you first start using the chatbot, you can use the pre-built dialogue set instead of your own data to build your service based on the scenario production guideline.
Understanding Language-specific Expressions with Little Data
NAVER CLOUD PLATFORM's chatbot engine helps the chatbot to understand various expressions by providing language-specific information, such as the parts of speech in Korean. Also, a reinforcement algorithm is applied to have the model understand the 'difference between (similar questions-different responses) pairs' and the 'difference between (different questions-similar response) pairs'. This algorithm helps to amplify utterance patterns through self-learning using the small amount of utterance example data.
What can I help you with today?
I am not a software developer but can I make a chatbot?
[Go to the Chatbot Start Guide]
I want to provide an audio chatbot. What can I do?
Easy link (CSR/CSS) with Clova API is provided so you can expand your service to include an audio chatbot.
[Go to the Clova API Start Guide]
I want to link my service to a messenger like LINE and KakaoTalk.
[Go to the Chatbot Channel Link Guide]
To help you operate your service flexibly in different situations, we offer a practical service and the corresponding pricing model.
· The chatbot service operates with the following conditions.
· API request: Fees are charged on the total number of requests.
· Dialogue model build: Fees are charged based on the number of dialogue training sessions.
(Dialogue training refers to carrying out natural language processing and machine learning for the chatbot to continue a conversation in a natural way. Fees are charged by adding up the number of training sessions by each domain.)
Chatbot service provides a basic price plan for those seeking practicality
and a standard price plan for those running large-scale services.
Small businesses needing a chatbot /
Those who want to develop and test a chatbot
|Dialogue model build||API request|
Dialogue model build: 10 calls/month (maximum)
API call: 1,000 calls/month (maximum)
Enterprises and small/medium sized service providers
|Dialogue training||API request|
|Dialogue model build: 20,000 KRW/build|
(Maximum monthly training cost per domain: 3,000,000 KRW)
|0 - 1,000 calls||Free|
|1,000 - 1,000,000 calls||3 KRW|
|Over 1,000,000 calls||2 KRW|
- You can select a service type (Basic/Standard) for each domain.
You cannot change the service type once the domain is created. However, you can use the Excel Import or Export feature to move the dialogue scenarios between domains.
- Additional charges are applied for API Gateway usage because the chatbot is called via API Gateway when linking to messengers or custom channels after service deployment.
- Additional charges are applied if you connect to the Clova API (CSR/CSS) to implement voice chatbot.