Starting from Scratch in Semantic Role Labeling: Early Indirect Supervision

A fundamental step in sentence comprehension involves assigning semantic roles to sentence constituents. To accomplish this, the listener must parse the sentence, find constituents that are candidate arguments, and assign semantic roles to those constituents. Where do children learning their first languages begin in solving this problem? To experiment with different representations that children may use to begin understanding language, we have built a computational model for this early point in language acquisition. This system, Latent BabySRL, learns from transcriptions of natural child-directed speech and makes use of psycholinguistically plausible background knowledge and realistically noisy semantic feedback to improve both an intermediate syntactic representation and its final semantic role classification. Using this system we show that it is possible for a simple learner in a plausible (noisy) setup to begin comprehending the meanings of simple sentences, when initialized with a small amount of concrete noun knowledge and some simple syntax-semantics mapping biases, before acquiring any specific verb knowledge.