The Greedy and Recursive Search for Morphological Productivity
This work addresses the challenge of how children learn productive morphological rules despite exceptions, offering a computational model that better mimics human acquisition than existing methods.
The authors tackled the problem of modeling morphological productivity acquisition in children by proposing a greedy search model that automatically hypothesizes and evaluates rules over a vocabulary, recursively subdividing when needed. The model, trained on child-directed input, replicated developmental patterns like German noun pluralization and produced nonce word responses more similar to humans than neural networks, despite using less data.
As children acquire the knowledge of their language's morphology, they invariably discover the productive processes that can generalize to new words. Morphological learning is made challenging by the fact that even fully productive rules have exceptions, as in the well-known case of English past tense verbs, which features the -ed rule against the irregular verbs. The Tolerance Principle is a recent proposal that provides a precise threshold of exceptions that a productive rule can withstand. Its empirical application so far, however, requires the researcher to fully specify rules defined over a set of words. We propose a greedy search model that automatically hypothesizes rules and evaluates their productivity over a vocabulary. When the search for broader productivity fails, the model recursively subdivides the vocabulary and continues the search for productivity over narrower rules. Trained on psychologically realistic data from child-directed input, our model displays developmental patterns observed in child morphology acquisition, including the notoriously complex case of German noun pluralization. It also produces responses to nonce words that, despite receiving only a fraction of the training data, are more similar to those of human subjects than current neural network models' responses are.