We have been testing Dialogflow as a team for the first time: All of us were given 3 tasks at the beginning to work with.
Csaba had to demo the food ordering. József had to demo the food tablebooking, and Márk had to demo the food order altering.
There has been several conculison/learnings to do, just to highlight the few:
Using contexts, as József has encountered is kind of troublesome. There were some interesting behaviour noticed.
Food ordering cannot really be done simply by using prompts, as some important inputs might depend on further information that has to be retrieved from the agent.
Food order altering can be extremely difficult when it comes to aggregated modifications in the order. e.g.: There is a street adress altering + size altering + a diet coke pls.
The question of "a" and "an" can be solved by entities or simply by coding.
in the webhook, the way it works is simply the following: The agent sends a JSON object to the designated URL or to the inline editor's stuff. There is plenty of data of in those.
See created issues for more details ( + doc about use cases/training phrases).
2018.10.23 - Dialogflow testing
We have been testing Dialogflow as a team for the first time: All of us were given 3 tasks at the beginning to work with. Csaba had to demo the food ordering. József had to demo the food tablebooking, and Márk had to demo the food order altering. There has been several conculison/learnings to do, just to highlight the few:
See created issues for more details ( + doc about use cases/training phrases).