Modeling morphology with Linear Discriminative Learning: considerations and design choices

TL;DR

This paper explores how Linear Discriminative Learning can be used to model German noun inflection, emphasizing the importance of representation choices, incremental learning, and context in capturing morphological patterns.

Contribution

It demonstrates how to effectively model German noun inflection using Linear Discriminative Learning, highlighting methodological considerations and the impact of representation and learning strategies.

Findings

Model accurately captures known German noun inflection patterns.

Incremental learning is essential for modeling frequency effects.

Model shows limited generalization to unseen data, reflecting semi-productivity.

Abstract

This study addresses a series of methodological questions that arise when modeling inflectional morphology with Linear Discriminative Learning. Taking the semi-productive German noun system as example, we illustrate how decisions made about the representation of form and meaning influence model performance. We clarify that for modeling frequency effects in learning, it is essential to make use of incremental learning rather than the endstate of learning. We also discuss how the model can be set up to approximate the learning of inflected words in context. In addition, we illustrate how in this approach the wug task can be modeled in considerable detail. In general, the model provides an excellent memory for known words, but appropriately shows more limited performance for unseen data, in line with the semi-productivity of German noun inflection and generalization performance of native German speakers.