We use the models adapted on [Multiwoz](https://www.aclweb.org/anthology/D18-1547) dataset to build our agent. This pipeline agent consists of NLU, DST, Policy and NLG modules.
We use the models adapted on [Multiwoz](https://www.aclweb.org/anthology/D18-1547) dataset to build our agent. This pipeline agent consists of NLU, DST, Policy and NLG modules.
First, import some models:
First, import some models:
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
# common import: convlab2.$module.$model.$dataset
# common import: convlab2.$module.$model.$dataset
from convlab2.nlu.jointBERT.multiwoz import BERTNLU
from convlab2.nlu.jointBERT.multiwoz import BERTNLU
from convlab2.nlu.milu.multiwoz import MILU
from convlab2.nlu.milu.multiwoz import MILU
from convlab2.dst.rule.multiwoz import RuleDST
from convlab2.dst.rule.multiwoz import RuleDST
from convlab2.policy.rule.multiwoz import RulePolicy
from convlab2.policy.rule.multiwoz import RulePolicy
from convlab2.nlg.template.multiwoz import TemplateNLG
from convlab2.nlg.template.multiwoz import TemplateNLG
from convlab2.dialog_agent import PipelineAgent, BiSession
from convlab2.dialog_agent import PipelineAgent, BiSession
from convlab2.evaluator.multiwoz_eval import MultiWozEvaluator
from convlab2.evaluator.multiwoz_eval import MultiWozEvaluator
from pprint import pprint
from pprint import pprint
import random
import random
import numpy as np
import numpy as np
import torch
import torch
```
```
%% Cell type:markdown id: tags:
%% Cell type:markdown id: tags:
Then, create the models and build an agent:
Then, create the models and build an agent:
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
# go to README.md of each model for more information
# go to README.md of each model for more information
That's all! Let's chat with the agent using its response function:
That's all! Let's chat with the agent using its response function:
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("I want to find a moderate hotel")
sys_agent.response("I want to find a moderate hotel")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("Which type of hotel is it ?")
sys_agent.response("Which type of hotel is it ?")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("OK , where is its address ?")
sys_agent.response("OK , where is its address ?")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("Thank you !")
sys_agent.response("Thank you !")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("Try to find me a Chinese restaurant in south area .")
sys_agent.response("Try to find me a Chinese restaurant in south area .")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("Which kind of food it provides ?")
sys_agent.response("Which kind of food it provides ?")
```
```
%% Cell type:code id: tags:
%% Cell type:code id: tags:
```
```
sys_agent.response("Book a table for 5 , this Sunday .")
sys_agent.response("Book a table for 5 , this Sunday .")
```
```
%% Cell type:markdown id: tags:
%% Cell type:markdown id: tags:
## Build a simulator to chat with the agent and evaluate
## Build a simulator to chat with the agent and evaluate
In many one-to-one task-oriented dialog system, a simulator is essential to train an RL agent. In our framework, we doesn't distinguish user or system. All speakers are **agents**. The simulator is also an agent, with specific policy inside for accomplishing the user goal.
In many one-to-one task-oriented dialog system, a simulator is essential to train an RL agent. In our framework, we doesn't distinguish user or system. All speakers are **agents**. The simulator is also an agent, with specific policy inside for accomplishing the user goal.
We use `Agenda` policy for the simulator, this policy requires dialog act input, which means we should set DST argument of `PipelineAgent` to None. Then the `PipelineAgent` will pass dialog act to policy directly. Refer to `PipelineAgent` doc for more details.
We use `Agenda` policy for the simulator, this policy requires dialog act input, which means we should set DST argument of `PipelineAgent` to None. Then the `PipelineAgent` will pass dialog act to policy directly. Refer to `PipelineAgent` doc for more details.
Now we have a simulator and an agent. we will use an existed simple one-to-one conversation controller BiSession, you can also define your own Session class for your special need.
Now we have a simulator and an agent. we will use an existed simple one-to-one conversation controller BiSession, you can also define your own Session class for your special need.
We add `MultiWozEvaluator` to evaluate the performance. It uses the parsed dialog act input and policy output dialog act to calculate **inform f1**, **book rate**, and whether the task is **success**.
We add `MultiWozEvaluator` to evaluate the performance. It uses the parsed dialog act input and policy output dialog act to calculate **inform f1**, **book rate**, and whether the task is **success**.
The combination modes of pipeline agent modules are flexible. We support joint models such as MDBT, TRADE, SUMBT for word-DST and MDRG, HDSA, LaRL for word-Policy, once the input and output are matched with previous and next module. We also support End2End models such as Sequicity.
The combination modes of pipeline agent modules are flexible. We support joint models such as MDBT, TRADE, SUMBT for word-DST and MDRG, HDSA, LaRL for word-Policy, once the input and output are matched with previous and next module. We also support End2End models such as Sequicity.
Available models:
Available models:
- NLU: BERTNLU, MILU, SVMNLU
- NLU: BERTNLU, MILU, SVMNLU
- DST: RuleDST
- DST: RuleDST
- Word-DST: SUMBT, TRADE, MDBT (set `sys_nlu` to `None`)
- Word-DST: SUMBT, TRADE, MDBT (set `sys_nlu` to `None`)
