Not quite there yet: Combining analogical patterns and encoder-decoder networks for cognitively plausible inflection

TL;DR

This paper explores combining analogical patterns with encoder-decoder networks to improve modeling human judgments on nonce lexeme inflection, demonstrating that such integration enhances predictive performance.

Contribution

It introduces models that incorporate analogical patterns into neural architectures, showing their effectiveness in replicating human-like inflection judgments without external resources.

Findings

Model 2 ranks second among submissions.

Analogical patterns improve prediction accuracy.

Endogenous models outperform baseline approaches.

Abstract

The paper presents four models submitted to Part 2 of the SIGMORPHON 2021 Shared Task 0, which aims at replicating human judgements on the inflection of nonce lexemes. Our goal is to explore the usefulness of combining pre-compiled analogical patterns with an encoder-decoder architecture. Two models are designed using such patterns either in the input or the output of the network. Two extra models controlled for the role of raw similarity of nonce inflected forms to existing inflected forms in the same paradigm cell, and the role of the type frequency of analogical patterns. Our strategy is entirely endogenous in the sense that the models appealing solely to the data provided by the SIGMORPHON organisers, without using external resources. Our model 2 ranks second among all submitted systems, suggesting that the inclusion of analogical patterns in the network architecture is useful in mimicking speakers' predictions.