Intent-based NLP is a technique we’ve beenusing quite recently on our projects, and its key difference is that thedeveloper has to train certain intents, which a Chatbot (for example) shouldhandle. For those intents, it needs to receive data, so that if there is a greet_intentto initiate the conversation with the end user, the data samples would looklike this: “[Hello, Welcome, Hi, Good Morning]”.
With GPT-3.5 and the upcoming GPT-4 models, wedon't have to do this anymore. All the model needs, is to detect a certainintention defined by the developer. The better described, the more reliable itwill be. This removes the need to provide training data; only an initial promptdescribing the intention.
With that said, here are other key advantagesof ChatGPT versus Intent-based NLP:
Overall, while intent-based NLP systems areuseful for certain types of applications, and especially powerful for business intelligence, ChatGPT's flexibility,natural language processing capabilities, ability to learn, and personalizationmake it a more versatile and powerful language model in a wide range ofcontexts. It’s important to note that if an intent-based NLP doesn’t supportgenerating human-like responses, it most likely means that Natural languagegeneration (NLG) wasn’t implemented, which does not have to be directly tied toall intent-based NLP systems.
SABOT is currently our company’s example of asuccessful intent-based NLP model. We will look into using LLM’s (LargeLanguage Models) such as ChatGPT to enhance its feature set and abilities evenfurther in the future. Learn more here.