Natural Language Understanding

Built-In Regularization

Simple chatbots match soley on keywords. If there isn’t a perfect match, the bot fails. Pez.AI chatbots use real machine learning and are regularlized. That means they can match on phrases that haven’t been seen before.

Spelling and Grammar

Users don’t need to have perfect spelling and grammar to interact with Pez.AI chatbots. They handle these variations just fine.

Relative Time Concepts

Pez.AI chatbots automatically translate relative time concepts, like “tomorrow”, into actual dates and times.

Named Entity Recognition

Named entities capture different classes of concepts, such as dates, cities, people. Named entity recognition is a technique for capturing these concepts within text. Pez.AI chatbots automatically detect named entities. The captured entities can be used within bot responses as well as in Pez expressions.

Sentence Type Detection

It’s important to know when a user asks a question versus makes a statement. We built a custom model to detect different sentence types, which helps provide conversational intelligence.