Chatbots Orchestration and Multilingual Challenges – Part 1

Share this post:

Chatbots Orchestration and Multilingual Challenges

Nowadays, conversational bots are created to meet the needs for users. However, if users need to actively switch between languages and conversations, or maintain a large conversational bot, it is also a tedious and difficult task, and it often causes bad experience. We propose a multilingual chatbot architecture to expand languages support. We also build a chatbots orchestration architecture, where professional questions are answered by professional bots, and questions that cannot be answered are directed other bots or knowledge bases, allowing chatbots to be independently maintained and collaborate to answer questions. This blog post will cover chatbots orchestration first then make multilingual chatbots in another post.

Part 1. Chatbots Orchestration

In the chatbots orchestration architecture, we leverage Watson Natural Language Classifier (NLC) to classify users’ questions. And the Classifier training data are from each chatbot’s intent and entities. It does not to be the whole set of intent and entities. Only the data that can help NLC to determine the category is good enough. With the output of the classifier, we can see the percentage of confidence of each chatbot in the backend.  Orchestrator then dispatches user’s utterance to corresponding chatbot.

So, in below example, a user asks a question, “Could you please inactivate my card ?”. The NLC calculates that it has 85% confidence to be classified as a Credit Card question.  Orchestrator takes the NLC result then directs user’s question to Credit Card bot to answer question. We can also add some rules in Orchestrator when directing questions.  ex: if none of the classification score is greater than 75%, then Orchestrator directs the question chitchat bot.

Advanced Chatbots Orchestration

This is an advanced architecture that could be applied in real customer support. The same architure, but we involve Tone analyzer to get user’s emotion in order to redirect to human customer service. Also if the confidence score from chatbot is not good, Orchestrator could redirect the question to Watson Discovery Server. This could handle the long tail question scenario, at least to show result from Discovery Server instead of bad answer from chatbots.

Advantages of Chatbots Orchestration

  • Maintain each specific domain chatbot and then orchestrate them easily
  • Orchestrator rules provides flexibility to have better questions dispatch. Both short tails questions (questions commonly asked) and long tails questions could be handled
  • Further integration with multilingual chatbots so that it could serve users in a worldwide company

Learn More

Ready to globalize your chatbots to reach global customers in their native language? Below are resources to help you:

Part 2. Multilingual Chatbots Challenges

We will continue to talk about multilingual chatbots challenges in this post. Let’s see how easily to expand a single language chatbot to be a multilingual one.


GSSC Chief Architect

More Watson stories
December 24, 2018

Chatbots Orchestration and Multilingual Challenges – Part 2

Chatbots Orchestration and Multilingual Challenges This post is the part 2 of “Chatbots Orchestration and Multilingual Challenges” post. Please check Chatbots Orchestration, if you haven’t. Part 2. Multilingual Chatbots Challenges There are some challenges when creating multilingual chatbots. A common practice is to use a baseline language chatbot then translate it to other languages chatbots. […]

Continue reading

December 11, 2018

Announcing the GA Version of Watson Compare & Comply

Watson Compare & Comply has gone GA! Extract data from governing documents and other complex unstructured documents to power business processes with Watson.

Continue reading