| 0:00:15 | hi everybody and then attention from cmu this the joint work done with my colleagues |
|---|
| 0:00:20 | and one vice the mixing and so they might talk is going to talk about |
|---|
| 0:00:23 | how to making a charity slot filling systems using the encoder decoder models |
|---|
| 0:00:30 | so the slot filling system is the focus of this talk and it has been |
|---|
| 0:00:34 | a device critically in recent years because we have better models more data |
|---|
| 0:00:39 | but the is the one limitation of a just a two-pass approach is most of |
|---|
| 0:00:45 | the proposed approach is sometimes of dialogue system is the with domain specific and it |
|---|
| 0:00:50 | is restricted to one type of dialogue act states and there is not very domain-dependent |
|---|
| 0:00:56 | so because of that we have difficulty of extending existing operating system |
|---|
| 0:01:00 | when you don't lengths on your skills |
|---|
| 0:01:03 | and a second difficulties sometime in a way deploy a system in real life this |
|---|
| 0:01:08 | always out-of-domain a request and it's gonna be very limited to handling those out-of-domain request |
|---|
| 0:01:16 | so our goal is we want to develop a domain general system that is not |
|---|
| 0:01:21 | restricted to a the define predefine acts of state even this task |
|---|
| 0:01:26 | and the encoder-decoder model things a promising a choice because it's a very powerful and |
|---|
| 0:01:32 | its prosody model and it has been successfully applied to translation and the open domain |
|---|
| 0:01:37 | training |
|---|
| 0:01:38 | so the idea is quite simple we have a encoder that encodes the dialogue history |
|---|
| 0:01:42 | and of generating a dialog state representation and we have a decoder that you called |
|---|
| 0:01:48 | the system responds spoken by token |
|---|
| 0:01:51 | i |
|---|