The Role of Interpretable Patterns in Deep Learning for Morphology

TL;DR

This paper explores how character pattern recognition within a modified sequence-to-sequence model can improve understanding of morphological analysis, lemmatization, and copying tasks across multiple languages.

Contribution

It introduces a pattern matching encoder that identifies important subwords for different tasks and compares their roles using a novel similarity metric across languages.

Findings

Patterns reveal task-specific subword importance.

Similarity scores show relationships between tasks.

Method enhances interpretability of deep models.

Abstract

We examine the role of character patterns in three tasks: morphological analysis, lemmatization and copy. We use a modified version of the standard sequence-to-sequence model, where the encoder is a pattern matching network. Each pattern scores all possible N character long subwords (substrings) on the source side, and the highest scoring subword's score is used to initialize the decoder as well as the input to the attention mechanism. This method allows learning which subwords of the input are important for generating the output. By training the models on the same source but different target, we can compare what subwords are important for different tasks and how they relate to each other. We define a similarity metric, a generalized form of the Jaccard similarity, and assign a similarity score to each pair of the three tasks that work on the same source but may differ in target. We examine how these three tasks are related to each other in 12 languages. Our code is publicly available.