Previous approaches rely on a pipeline of individually optimized components containing state tracking, dialog policy/management etc requiring expensive annotations and are hard to scale.
-
-
Prikaži ovu nit
-
In our experiments we find that: 1) our model was able to incorporate external knowledge and generate factual text response with weak supervision signal. 2) our model can incorporate medium-size knowledge bases with only 8K training examples over multiple verticals.pic.twitter.com/TjFvXO9e4G
Prikaži ovu nit -
The model is trained at turn-level where the dialog history fed into model as input has previous ground-truth turns of the dialog. In the conversations here the actual text responses generated by model itself are used as the assistant’s side of dialog history to be fed as input.pic.twitter.com/REdOXsnMYT
Prikaži ovu nit -
Prikaži ovu nit
-
Implementation of Neural Assistant: Joint Action Prediction, Response Generation, and Latent Knowledge Reasoning:https://github.com/tensorflow/tensor2tensor/blob/master/tensor2tensor/models/neural_assistant.py …
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
Exciting. How does it handle queries with timely information like "find me an inexpensive Italian restaurant in San Francisco for for people tonight?"
-
thanks! IIUC your question is about how to incorporate the information whether the restaurant has seats for tonight or not. I don't think MultiWOZ dataset has that info, but please check out our Taskmaster dataset (https://ai.google/tools/datasets/taskmaster-1 …)
- Još 2 druga odgovora
Novi razgovor -
-
-
Wow this is really neat! Now how will we be able to integrate this into a Google Assistant Action? I think there's a good opportunity to create great conversational demos.
-
Thanks for the interest. I think Neural Assistant + Taskmaster (https://ai.google/tools/datasets/taskmaster-1 …) + Google search results as source for external knowledge can work really well for task-oriented dialog!
Kraj razgovora
Novi razgovor -
-
-
Awesome work!
-
Thanks! :)
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.