The JennyStudio uses artificial intelligence to analyze user’s messages and to find the most relevant reply.
Understanding the Scoring
Queries and Keywords are determined in each Default state. This how you predict what the customers can ask from the chatbot. The closer the customer's message is to the queries and keywords you have taught the bot, the better the scoring is and the more accurate the reply is. Teaching several slightly different queries increases the probability of the state being returned.
The minimum score to return a state is 0.8. This is also called a 'threshold'. The maximum score with queries is 1, but can be even more when using keywords.
Scoring with Queries
Let's say we have a state called 'How are you?' which is conveniently named to answer questions when the customer is interested in how our chatbot is doing. The state takes in a few queries:
Now let's say the customer asks 'How are you doing ms. Chatbot?'. The score is not going to be 1 as we do not have that exact query taught to the chatbot. But this is completely fine, as we have taught 'how are you doing' as a query.
In this example our 'How are you?' state got a score of 0.89.. and no other states got a score of more than 0.8. If there are multiple states that get scores of more than 0.8, chatbot will return the state that got the highest score.
If there are no states that get a score of more than 0.8, chatbot will return a Fallback state.
Scoring with Keywords
With Keywords your score can climb even over 1! I modified the same state to accept just keywords this time:
As there is the word 'and' between the keyword groups, ALL of these groups must match to the query by the customer. When all of them match, we get a score of over 1!
As you can see, Keywords and especially Keyword Groups produce a higher score but this will require you to predict more accurately what the customer might ask.