MIXED PROPOSAL BASED MODEL TRAINING SYSTEM

摘要

In implementations of the subject matter described herein, each token for containing an element in the training data is sampled according to a factorization strategy in training. Instead of using a single proposal, the property value of the target element located at the token being scanned is iteratively updated one or more times based on a combination of an element proposal and a context proposal. The element proposal tends to accept a value that is popular for the target element independently of the current piece of data, while the context proposal tends to accept whenever the property value that is popular in the context of the target data or popular for the element itself. The proposed modeling training approach can converge in a quite efficient way.

主权项

1. A device comprising:
a processing unit; a memory coupled to the processing unit and including modules, when executed by the processing unit, cause the device to sample a token in target data in a corpus that includes multiple pieces of data, the modules including:
an element proposal determining module configured to determine an element proposal associated with a first candidate value for a property of a target element located at the token, the element proposal indicating a probability that the first candidate value is assigned to the property independently of a context of the target data;a context proposal determining module configured to determine a context proposal associated with a second candidate value for the property of the target element, the context proposal indicating a probability that the second candidate value is assigned to the property in the context of the target data; anda property updating module configured to update a current value of the property of the target element based on a combination of the element proposal and the context proposal.