The Greedy and Recursive Search for Morphological Productivity

TL;DR

This paper introduces a greedy recursive search model for morphological productivity that mimics child language acquisition, effectively handling exceptions and generalizing to new words, outperforming neural networks in nonce word responses.

Contribution

It proposes a novel greedy recursive search algorithm for hypothesizing and evaluating morphological rules, improving upon previous methods that require fully specified rules.

Findings

Model reproduces developmental patterns in child morphology acquisition.

Successfully generalizes to complex cases like German noun pluralization.

Outperforms neural network models in nonce word response similarity.

Abstract

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.