Lab 2,3,4,5 submission #4
Conversation
Also add report.
|
Comments on lab 2 submission: Good job. If you go through all the steps to create a meeting, but answer "no" to the final question confirming the details, then something strange happens: Do you want me to create a meeting with ... on ... -> No -> Will it take the whole day? -> No -> Will it take the whole day? -> ... My best guess is that that the slots for person and day have already been filled, so the system fast-tracks past these and tries to get a new time only. |
vladmaraev
force-pushed
the
main
branch
4 times, most recently
from
f89ca27 to
96cbfe4
Compare
|
Lab5 comments: OK. The confidence thresholds as you've implemented them seem to be working pretty well, but the NLU threshold is of course more interesting when you use a trained NLU model. The hard-coded slot filler "alex on monday" leads to a weird interaction, with the system asking "Will it take the whole day?" while it is in reality listening for a time. Regardless of whether I answer "yes" or "no" it lets me know that it didn't understand and asks me to repeat the time. |
No description provided.