[ad_1]
And while this process seems straightforward to the user, Amazon insists that the actual backend algorithms that handle the functionality are quite complicated. The trigger that activates the system and provides a latent goal – set the reminder alarm or turn on the DVR – is trained on a deep learning model that takes into account several aspects of the context of the conversation. Additionally, the model will improve over time as it associates more contextual clues with latent goals based on its ongoing conversations with users.
The system improves by actively learning, “which identifies examples of interactions that would be particularly instructive in future adjustments,” according to a press release Tuesday. The system also searches for “named entities and other arguments in the current conversation”, tagging and automatically formatting them for use by other Alexa skills. The most specific and useful recommendations are implemented while the least are discarded, a process known as learning from bandits.
It is currently available in English for the United States so developers can start implementing it immediately.
[ad_2]
Source link