Check out the following examples to understand the differences between keywords and machine learning matching systems.
The bot can enter one of the following interactions.
All examples are calculated with the default confidence score (0.7).
- Interaction A uses Machine Learning matching system: The section User Says is filled with:
Tell me more about your payments policy.
- Interaction B uses Keywords matching system: The section User Says is filled with one word
Now let’s see the bot’s behavior with three different user inputs.
payments- would trigger Interaction B.
- Matching score: ML < 70%; Keywords = 100%.
- Explanation: The matching score for interaction B equals 100% while for interaction A the matching score is less than 70% (below the confidence score).
tell me about payments policywould trigger interaction A.
- Matching score: ML => 70% but it’s < 100%; Keywords: 100%
- Explanation: Even though the matching score for the interaction A is lower, ML has preference and will be executed.
payments policy: would trigger Interaction B
- Matching score: ML < 70%, Keywords = 100%;
- Explanation: The matching score for machine learning is below the required threshold. Interaction A won’t be considered so the bot will go for the Interaction B.
ℹ️ Even though ML matching system has a preference, it won’t be selected when the matching score Is lower than the confidence score.
We’ve got two following interactions:
- Interaction A - based on ML matching engine. User Says section is filled with
- Interaction B - based on keywords matching engine. User Says section is filled with a user entity named
Movies. The entity has two values:
Now let’s check how is our bot going to behave after the following user input is added:
- Matching score: ML = 100%, Keywords = 0%; triggered Interaction A
sys.anyIs the only possible entity what won’t have preference while powered by the ML matching system. Interaction B matching score equals 0%. The bot selects Interaction A as this is the only possible variant.
- Matching score: ML = 100%, Keywords = 100%; triggered Interaction B
- Explanation: Having in mind that
sys.anyIs never prioritized, the bot will follow Interaction B which has matching score 100%.
🦉 Machine Learning matching system isn’t prioritized when User Says field is filled with the system entity
Let’s see the last example. We’ve got two interactions:
- Interaction A - based on ML matching system. User Says filled with a phrase
- Interaction B - based on Keywords matching system. User says filled with a phrase
Both interactions have identical User Says field. Let’s check how will this scenario go.
- Matching score: ML > 70%, Keywords = 100%; triggers Interaction A
- Explanation: ML interaction A is prioritized as it’s ML > 70%.
- Matching score: ML = %, Keywords = 0%; triggers Fallback Interaction If ON.
- Explanation: No condition has been met.
Interactions can have vary matching systems. Select the one that is more relevant and gives bigger chances to be selected by the bot.