Today we discuss a chatbot task related to intent detection–entity extraction. Entity extraction is classified in the Training category of chatbot product features. It refers to the chatbot’s capability of recognizing specific keywords and key phrases from the customer’s inputs and using them along with the detected intent to execute the relevant task or deliver the appropriate response.
Most chatbot building platforms provide two types of entities–system entities and named entities. System entities are generic keywords such as date, time, city, country–and are usually automatically configured and detected by the chatbot platform. Named entities, on the other hand, are custom entities specific to the use case and configured by the enterprise user when developing the chatbot, such as “Destination City,” “Payment Date,” and “Account Number.”
For example, while training the utterance “I would like to book a flight from San Francisco to London on June 16,” the user would set “San Francisco” as the “Source City” named entity, set “London” as the “Destination City” named entity, and set “June 16” as the “Flight Date” named entity. After having trained the chatbot with a sufficient number of utterances and configured the named entities within, the chatbot could then be deployed to understand customer inputs and extract the entities in them. If the customer input is “Please book me a flight from Los Angeles to Paris on July 30,” the chatbot would recognize “Los Angeles” as the “Source City” named entity, “Paris” as the “Destination City” named entity, and “July 30” as the “Flight Date” named entity. It would then check for available seats on such flights in its backend database and take the customer to the next step in the flight booking process.
The screenshot above shows a list of entities in an example chatbot vendor – Voiceflow.