I am creating a chatbot using Google DialogFlow, which includes several pages. These pages are uploaded as pdf file using Knowledge (Beta) feature in Google DialogFlow. During the testing of chatbot response, it is observed that sometimes the chatbot response is inaccurate. Here is an example to describe this problem
Question 1 (format-1): when inventory structure should be planned? Reply: You can plan your inventory structure before creating inventory in Ad Manager to get the most out of your ad space on web or app. #inventory
Question 1 (format-2): should I plan inventory structure after or before creating inventory in ad Manager? Reply: You can plan your inventory structure before creating inventory in Ad Manager to get the most out of your ad space on web or app. #inventory
Question 1 (format-3): should I plan inventory structure after creating inventory in ad Manager? Reply: Your inventory is a combination of ad units, placements, and key-values that allows you to target horizontally across your content. The examples below show how you might use ad units, key-values, and system-defined criteria to meet inventory requirements.
As show above, the response from the chatbot is correct for the Question 1 (format-1) and Question 1 (format-2). However, the chatbot's response is incorrect for the Question 1 (format-3).
I do not want to add the training phrases provided in the Intent feature available within Google Dialogflow to improve chatbot accuracy as this would require training for each question. Is there a way to improve chatbot accuracy without using trarining phrases?
