Predicting functional tags of semantic role labeling

A computer program product for predicting functional tags of semantic role labeling, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

• receive, at a scheduler utilizing the processor, a training data instance and a target instance;
• generate, by the scheduler utilizing the processor, an input sequence from the training data instance and the target instance;
• send the input sequence from the scheduler to an encoder, utilizing the processor;
• map, by the encoder utilizing the processor, the input sequence to a feature vector;
• send the feature vector from the encoder to the scheduler, utilizing the processor;
• send the feature vector from the scheduler to a predictor, utilizing the processor;
• and
• map, by the predictor utilizing the processor, the feature vector to a class vector to create a label for the target instance.

A system, comprising:

• a processor;
• and
• logic integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:
• receive, at a scheduler, a training data instance and a target instance;
• generate, by the scheduler, an input sequence from the training data instance and the target instance;
• send the input sequence from the scheduler to an encoder;
• map, by the encoder, the input sequence to a feature vector;
• send the feature vector from the encoder to the scheduler;
• send the feature vector from the scheduler to a predictor;
• and
• map, by the predictor, the feature vector to a class vector to create a label for the target instance.

A computer-implemented method, comprising:

• receiving training data, wherein each training datum includes a dependency tree having zero or more sub-trees labeled with ground-truth functional tags;
• training a model that includes a Scheduler, an Encoder, and a Predictor, wherein:
• (i) the Scheduler is any function that maps a feature vector to an action ranking vector, determines tokens to pass to the Encoder, and decides when to invoke the Predictor;
• (ii) the Encoder is any function that maps a sequence of input data to a feature vector;
• (iii) the Predictor is any function that maps a feature vector to a vector of functional tags;
• and
• (iv) wherein during the training, an error of prediction is used to adjust parameters of the Predictor, the Encoder, and the Scheduler;
• and
• applying the trained model to predict functional tags in a sentence.