🔍 Rule-Based Syntax for Intents and Entities

Rule-based syntax is a simple, word-matching approach to triggering an intent or entity without relying on an NLP model.

Matching Basics

• A sample matches any word that begins with the same characters as the sample.

  Example:

  • Sample al matches aloha and alcohol, but not banal.

• A multiple-choice sample provides alternate options and is expanded into individual matches.

  Example:

  • Sample Alibaba/Aladdin matches both Alibaba and Aladdin.

Handling Multiple Matches

When there are multiple matches for a given entity type:

1. Wrapper Handling:

   • If the entity has a wrapper, it determines which match to keep.

2. First Match Preference:

   • If no wrapper is used, the first match in the message is kept.

   Example:

   • If the sample is Al and the message is "Alibaba and Aladdin," both match, but Alibaba is chosen since it appears first.

3. Matches Starting at the Same Position:

   • If two samples start at the same position, conflicts arise. For example, if the entity type has samples Al (for Aladdin) and Alice, then both match Alice.

   Solution:

   • Use - to exclude specific matches. For instance, use Al -Alice to match Aladdin and exclude Alice.

   ❗️

   Caution

   If someone says "I want Aladdin, not Alice," you might still match Alice.

Syntax Recap

• " / ": Separate multiple-choice options (e.g., Alibaba/Aladdin)
• " - ": Exclude specific matches (e.g., Al -Alice)
• " [ ] ": Enclose exact expressions (e.g., [Hello World])