BPE vs. Morphological Segmentation: A Case Study on Machine Translation of Four Polysynthetic Languages

TL;DR

This study compares morphological segmentation methods and Byte-Pair Encodings for machine translation of four polysynthetic languages, revealing that unsupervised morphological segmentation often outperforms BPEs in translation quality.

Contribution

The paper introduces a comprehensive comparison of segmentation methods for polysynthetic languages and provides new datasets and a parallel corpus for Raramuri and Shipibo-Konibo.

Findings

Unsupervised morphological segmentation outperforms BPEs in most language pairs.

Supervised segmentation achieves better scores but underperforms in translation tasks.

New datasets and a parallel corpus for Raramuri and Shipibo-Konibo are provided.

Abstract

Morphologically-rich polysynthetic languages present a challenge for NLP systems due to data sparsity, and a common strategy to handle this issue is to apply subword segmentation. We investigate a wide variety of supervised and unsupervised morphological segmentation methods for four polysynthetic languages: Nahuatl, Raramuri, Shipibo-Konibo, and Wixarika. Then, we compare the morphologically inspired segmentation methods against Byte-Pair Encodings (BPEs) as inputs for machine translation (MT) when translating to and from Spanish. We show that for all language pairs except for Nahuatl, an unsupervised morphological segmentation algorithm outperforms BPEs consistently and that, although supervised methods achieve better segmentation scores, they under-perform in MT challenges. Finally, we contribute two new morphological segmentation datasets for Raramuri and Shipibo-Konibo, and a parallel corpus for Raramuri--Spanish.