AI System to Determine Actionable Intent
AssigneeMicrosoft Technology Licensing, LLC
InventorsChu-Cheng Lin, Michael Gamon, Dongyeop Kang, Patrick Pantel, Madian Khabsa, Ahmed Hassan Awadallah
Life of this patent
Estimating the term…
Abstract
A method includes receiving an email addressed to a recipient user, processing the received email using a reparametrized recurrent neural network model to identify an action based on the received email, and wherein the reparametrized recurrent neural network model has been trained on an email dataset annotated with recipient corresponding actions and reparametrized on unannotated conversation data having structures similar to email data.
Claims20
A method comprising:
- receiving an email addressed to a recipient user;
- processing the received email using a reparametrized recurrent neural network model to identify an action based on the received email;
- and
- wherein the reparametrized recurrent neural network model has been:
- trained on an email dataset annotated with recipient corresponding actions;
- and
- reparametrized on unannotated conversation data having structures similar to email data.
A machine-readable storage device having instructions for execution by a processor of a machine to cause the processor to perform operations to perform a method of generating a model, the method comprising:
- receiving a communication addressed to a recipient user;
- processing the received communication using a reparametrized recurrent neural network model to identify an action based on the received email;
- and
- wherein the reparametrized recurrent neural network model has been:
- trained on a communication dataset annotated with recipient corresponding actions;
- and
- reparametrized on unannotated conversation data having structures similar to the communication dataset data.
A computer implemented method of training a recurrent neural network (RNN) for identifying actions responsive to received emails, the method including operations comprising:
- initializing weights of the RNN;
- training the RNN on a target domain set of recipient action annotated emails;
- and
- reparametrizing the RNN on unannotated conversation data threads.
The computer implemented method of claim 34 wherein the operations further comprise:
- evaluating the RNN on a validation set of action annotated emails to determine an accuracy of classification;
- repeating the method of claim 1 until a predetermined number of iterations is reached;
- and
- if the determined accuracy is the highest accuracy in the iteration, update the best parameters so far.
Dependencies are inferred from the claim text.