From the course: Artificial Intelligence for Business Leaders
Natural language processing (NLP)
From the course: Artificial Intelligence for Business Leaders
Natural language processing (NLP)
- One of the most interesting areas in artificial intelligence is natural language processing, or NLP. NLP allows artificial intelligence systems to both read and write using natural language. The system can read using natural language understanding, or NLU. Then the system writes using natural language generation, or NLG. Natural language processing can also involve speech recognition, which is when a machine identifies spoken words and converts them to text. But NLP doesn't need speech recognition. A lot of it is just about reading and writing text. So think of it this way. If you pick up your smartphone and ask the virtual assistant a question, it will actually break that down into several steps. Say you ask, "What's the weather today?" The first step is for the smartphone to take your audio file and upload it to a speech recognition system. The next step is then to convert that audio into text. Once your smartphone has the text, the last step is to use natural language understanding to convert it into something that a search engine might understand. So it could pull out key words such as "weather", and translate "today" into "current conditions". After it's gone through this NLU process, then it will use natural language generation. This will output the results using friendly text, instead of a chart or report. So instead of just having a picture with the current weather, your smartphone might respond, "It will be sunny and 73 today." So to you, it might seem like a simple process. You ask a question about the weather, and you get an appropriate response. But to the AI system, the natural language processing has undergone all three steps. Now, because there are so many smartphones, speech recognition has become extremely popular. But organizations often use NLP for email and online chats. For example, AI systems could use natural language understanding to read incoming messages. So that means, if a customer sent an email to your organization, it could automatically be routed to the correct person. Then organizations could use natural language generation to send out an appropriate response. Maybe the customer will instantly be notified that your organization understands their challenges. Then their issue will be routed to the correct person. You could also use natural language generation systems to explain complex reports, or even create news articles. So even though it might seem like speech recognition presents some of the greatest challenges, it's actually natural language understanding and generation that could be much more difficult. It's processing much more complex information. Language is filled with structural inconsistencies, and many people exaggerate or use metaphors. It's very difficult for an AI system to understand what you mean when you say you've been waiting forever, or that you're deeply disappointed. And you wouldn't want the system to send back a message that says something like, "I'm glad you're enjoying your product. Our support team will get back to you as soon as possible." Not only does the AI system need to figure out what you mean, the system has to come up with an appropriate response. That's the best way to ensure a positive experience for your customers.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.