@@ -21,11 +21,11 @@ The project is designed to easily integrate other models, and the code is writte
...
@@ -21,11 +21,11 @@ The project is designed to easily integrate other models, and the code is writte
## One-Shot Approach
## One-Shot Approach
The one-shot approach takes the MultiWOZ dataset to extract domain knowledge and use it as context for the dialogue generation. Then we prompt the model again with the generated Dialogue to create Annotations. The code for this approach is located in the [`src/one_shot`](src/one-shot/) directory.
The one-shot approach takes the MultiWOZ dataset and extracts domain knowledge to use as context for the dialogue generation. A Model is prompted to generate Dialogues and after that Annotations for these Dialogues. The code for this approach is located in the [`src/one_shot`](src/one-shot/) directory.
## Multiagent Approach
## Multiagent Approach
(TODO)
The Multiagent approach uses a retrieval-based system to generate dialogues. The system uses the HeidelKBerg dataset to retrieve information about restaurants, hotels and attractions in Heidelberg. The code for this approach is located in the [`src/multiagent`](src/multiagent/) directory.