metadata
language: en
tags:
- dstc10
- knowledge cluster classifier
widget:
- text: >-
oh and we'll mi thing uh is there bike clo ars or bike crac where i can
park my thee
- text: >-
oh and one more thing uhhh is there bike lockers or a bike rack where i
can park my bike
- text: >-
ni yeah that sounds great ummm dold you have the any idea er could you
check for me if there's hat three wifie available there
- text: >-
nice yeah that sounds great ummm do you have any idea or could you check
for me if there's uhhh free wi-fi available there
- text: perfect and what is the check kin time for that
This is the model used for knowledge cluster classification for the DSTC10 track2 knowledge selection task, trained with double heads, i.e., classifier head and LM head using ASR error simulator for model training.
For further information, please refer to https://github.com/yctam/dstc10_track2_task2 for the Github repository. You can use this model and use our source code to predict knowledge clusters under ASR errors. AAAI 2022 workshop paper: https://github.com/shanemoon/dstc10/raw/main/papers/dstc10_aaai22_track2_21.pdf