Conversational Language Understanding – Fundamentals of Natural Language Processing

Conversational Language Understanding

Conversational Language Understanding, or CLU, is a conversational AI service that is mostly used in bots to get relevant information from what users say (natural language processing). The CLU analysis includes two projects: dialogue and orchestration. You can use the “conversation” project if you want to pull out intents (the meaning behind a user’s words) and custom entities. To get the best response, you can also use the “orchestration” project, which brings together different language apps like Question Answering, Luis, and Conversation.

As the scenarios where AI is used get more complicated, people are using their voices to talk to digital apps more and more. In some cases, the conversations with AI agents can even turn out to be humanlike, approving them to be of good use for customer support apps and home automations.

Computers should not only be able to read what is typed into them, but they should also be able to translate it while keeping its semantic roots, so that the input doesn’t lose any of its real meaning. The Azure Conversational Language Understanding Service supports the comprehension of language. To use this service, the user has to understand three concepts: utterances, entities, and intentions.

Utterances

All the words a user might say that our system has to listen to and understand are called “utterances.” For instance, a user of a home automation system would say, “OK Google.”

• “Show me the picture of Golden Gate Bridge.”
• “Turn off the music.”

Entities

A subject to which our speech relates is called an “entity” for our system. Examples include “Golden Gate Bridge” and “music” in the following sentences:

• “Show me the picture of Golden Gate Bridge.”
• “Turn off the music.”

Intents

A direction or the final aim in the user’s speech is called an intent. For example, both of the previous commands have an intent. This means that we can build a system in our language interpreter that can wrap up and summarize these intentions.

A model of intentions and entities is defined by a language understanding program. For a given input, the model is trained to predict the right intent and the entities it will be applied to by using phrases.

Intents and entities are used to build a model, while utterances connect the two and tell the language understanding application what to do with the entity based on the intent.