NLP; NLU and NLG Conversational Process Automation Chatbots explained

Did you ever come across one of these terms during a research on Machine Learning or Chatbots? NLP, NLU, and NLG?

With all the buzz around Artificial Intelligence recently, I thought it would be useful to identify some of the core building blocks of Conversational AI while giving an example of its application in botique.ai Conversational Process Automation Chatbot Platform for Contact Centers.

NLP, NLU, and NLG all have one thing in common: They are always thrown around while talking about Conversational AI and Chatbots, and for many people, it usually seems that they’re all interchangeable with each other!

But don’t confuse them yet, it is correct that all three of them deal with human language, but each one is involved at different points in the process and for different reasons.

This quick article will try to give a simple explanation and will help you understand the major difference between them, and give you an understanding of how each is used.

What is the difference between NLG, NLP, and NLU?

These are the three basic acronyms used most frequently when discussing language related AI-technology:

• NLP — Natural Language Processing
• NLU — Natural Language Understanding
• NLG — Natural Language Generation

Employing these technologies builds conversationally intelligent applications that are able to engage with humans on a conversational level. Let us explore each one, and a real-life application will follow shortly!

Natural Language Processing (NLP)

NLP is the used to describe the ability of a system or a machine to ingest what is said, break it down, comprehend its meaning, determine appropriate action, and respond back in language the user will understand, according to Wikipedia “Natural Language Processing is a subfield of Artificial Intelligence that is concerned about the interaction between computers and human natural language”. NLP makes it possible for computers to read text, hear speech and interpret it, measure sentiment and even determine which parts are relevant. It has become really helpful resolving ambiguity in language and adds numeric structure to the data for many downstream applications.

Some of the basic NLP tasks are parsing, stemming, part-of-speech tagging, language detection and identification of semantic relationships. If you ever diagrammed sentences in primary school then you have done this manually before.

In general, NLP falls under the larger umbrella of Artificial Intelligence (AI), and it (NLP) combines both NLU & NLG as part of it.

Natural Language Understanding (NLU)

NLU is a subtopic of Natural Language Processing that uses AI to comprehend input made in the form of sentences in text or speech format. It enables computers to understand commands without the formalized syntax of computer languages and it also enables computers to communicate back to humans in their own languages.

NLU goes beyond just understanding the words, it interprets meaning in spite of human common human errors like mispronunciations or transposed letters or words. The main purpose of NLU is to create chat and speech-enabled bots that can interact effectively with a human without supervision.

Natural Language Generation (NLG)

The task of NLG is to generate natural language from a machine-representation system such as a knowledge base or a logical form. To simplify this, NLG is like a translator that converts data into a “natural language representation”, that a human can understand easily.

Some of the most common implementations are; written analysis for business intelligence dashboards, reporting on business data/data analysis, personalized customer communications via email and in-app messaging, IoT device status maintenance reporting, individual client financial portfolio summaries and updates, e-commerce product descriptions, and category landing page content.

How do NLP; NLU and NLG relate with each other?

Architecture Diagram for Chatbots ( https://goo.gl/3fRAvH )

When NLP breaks down a sentence, the NLU algorithms come into play to decipher its meaning. It is quite possible that the same text has various meanings, or different words have the same meaning, or that the meaning changes with the context.

Knowing the rules and structure of the language, understanding the text without ambiguity are some of the challenges faced by NLU systems. NLG does exactly the opposite; given the data, it analyzes it and generates narratives in conversational language a human can understand.

To build machines that understand natural language, we need to use NLP to perform the Sentiment analysis or opinion mining; which deals with using automatic analysis to find sentiments, emotions, opinions, and attitudes from a written text towards a subject. sentiments must be extracted, identified, and resolved, and semantic meanings are to be derived within a context and are used for identifying intents.

• Intents* activities that a user intends to do, or verbs (activities that the user needs to do). If we want to capture a request, or perform an action, use an intent. Like I want to purchase a data package, or I want to cancel my subscription.
• Entities* are the nouns or the content for the action that needs to be performed. In this example above “data package” and “subscription”.

For example, one use cases of botique.ai’s Digital Agent is the case of Travel Insurance. botique.ai Digital Agent receives a simple text from the customer wanting to purchase a Travel Insurance Coverage:

User: “I want to purchase insurance to Berlin starting tomorrow till Dec 9.”

Using Proprietary NLP Algorithms, the Digital Agent breaks down the sentence to understand it as follow:

purchase {intent} / Insurance {Entity} / Berlin {Entity} / Tomorrow (Nov 28) {Entity} / Dec 9 {Entity}

Using the geographical context and the current location of the user to set the destination details: Berlin, using the Time Context it to understand that Start Date “Tomorrow” and the End Date: Dec 9, 2018,

The platform is able to understand the request of the user, a Travel Insurance Package to Berlin from Nov 28 — Dec 9. The platform can verify further information like Age, Email, etc… to best decide the package. Request verification information like Account ID or password (or Two-way authentication). Connect to the enterprise system to provide the user with a price quote, user can proceed with payment, where the platform can verify the payment details and proceed with the purchase. Full Conversational Process Automation, without any human interaction.

Customer service is reaching a pivotal point where businesses must either embrace the technological tools of the future or risk their reputation by falling behind competitors. Either way, it will no longer be an option to stay the status quo. Exciting advancements are being made in this field. Follow botique.ai to stay up to date!