botpress/duckling

Duckling is a tool that helps software understand pieces of natural language text. When someone types "next Tuesday at 3pm" or "five kilometers," a computer sees only a string of characters. This project provides a system for identifying those phrases and converting them into structured data a calendar app or tracking tool can actually use. At a high level, it works by applying a set of composable, testable rules to incoming text. Instead of hard-coding logic to recognize every possible way someone might write a time or a distance, you build up small rules that can be combined. You string those rules together to evaluate whether an input string matches a pattern, then extract the meaningful information from it. The primary use case is chatbots and virtual assistants. If you are building a scheduling assistant, it needs to parse dates, times, and durations from free-text user messages. Someone might write "in a week," "tomorrow afternoon," or "July 4th." A developer would use this to recognize all those variations and turn them into standard dates that a backend database can process. The same logic applies to currency amounts, email addresses, or physical quantities. Notably, the project is written in Haskell, a programming language that enforces strict rules around data types and logic. That choice makes the system reliable and mathematically predictable when processing potentially messy text. The README itself doesn't go into further detail about specific setup or advanced features, but the core concept centers on providing a dependable foundation for language parsing.