ChatBot.com
keyboard_backspace

Features | 3 MIN READ

Big update of the keywords matching system

ChatBot.com

dariusz zabrzenski

on

ChatBot offers two matching systems: machine learning, and keywords. Each of them was designed to suit different needs. The right system choice can significantly help your chatbot to understand your clients, and increase the user satisfaction. One of them gets a vital update that brings better control over the matching process. Let’s talk about freshly updated keywords.

How keywords worked before the update

Machine learning analyses full queries and matches with user says, if possible. The system uses advanced algorithms to make this process smooth and user-friendly. Keywords, on the other hand, work excellent when we want a unique phrase or a word to trigger an interaction.

When we introduced keywords matching system we wanted our users to flag essential words and instruct the bot: “Hey! If you see this word, fire up the interaction!”. How simple is that? The action was pretty binary as we didn’t want to complicate the system. Keywords were matched with a tolerance set in the confidence score settings, regardless the distance between the words.

How does it work now

Updated keywords are more predictable and logical. First of all, we’ve decided the distance between two keywords. Now, the system matches keywords only if the distance between two keywords isn’t bigger than one word. We’ve also stripped the confidence score so the word in the query must be identical to the searched keyword. Last but not least, punctuation at the end of the word is allowed, e.g., pricing, as it doesn’t affect the meaning of the word.

Example

The best way to explain the behavior of updated Keywords is to use an example. Imagine that you own a small food truck business, and your bot takes food orders. One of your most famous menu options is a hot dog.

With the updated keywords system we can take orders in a more precise way. Let’s check out some of the cases!

Now let’s see if the marching is successful using various user input:

  • The client types: I would like to have a hot dog?
    ✅ This is perfect matching as our bot could find bot keywords.

  • The client types: Can I have dog hot?
    ✅ That’s pretty clumsy way to make an order but yes, our bot understands queries in reverse order.

  • The client types: Can I have hot a dog?
    ✅ The bot accepts maximum one word between our keywords. The bot takes the order.

  • The client types: Can I get dog jdhgf hot?
    ✅ Probably a cat run through the keyboard. The bot will understand the query as there’s one allowed word between keywords and the bot accepts reversed order.

  • The client types: Can I get a hut dog?
    ❌ The order won’t be taken. There’s a typo in the word hot.

  • The client types: Can I come with my dog. It’s really hot outside.
    ❌ Too many words between our keywords. The bot won’t recognize the query.

  • The client types: Can I have a hot dog, I’m so hungry.
    ✅ The bot will take the order. The punctuation at the end of the keyword is accepted.

  • The client says: Can I have a hot-dog?
    ❌ The bot won’t recognize these keywords. Two words with a punctuation mark in the middle are considered as one word.

Summary

Keywords matching system just got better, but it’s you who decides if it’s also better for your interactions. Remember to visit our Help Center to find out more about ChatBot’s artificial intelligence and know which system is more suitable for your needs.

comments powered by Disqus