Article

Beyond Characters: Subword-level Morpheme Segmentation

19th SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology

Peters, Ben; Martins, André FT — 2022 — Association for Computational Linguistics

Key information

Authors:

Peters, Ben (Benjamin Paul Oscar Peters); Martins, André FT (André Filipe Torres Martins)

Published in

July 14, 2022

Abstract

This paper presents DeepSPIN’s submissions to the SIGMORPHON 2022 Shared Task on Morpheme Segmentation. We make three submissions, all to the word-level subtask. First, we show that entmax-based sparse sequence-tosequence models deliver large improvements over conventional softmax-based models, echoing results from other tasks. Then, we challenge the assumption that models for morphological tasks should be trained at the character level by building a transformer that generates morphemes as sequences of unigram language model-induced subwords. This subword transformer outperforms all of our character-level models and wins the word-level subtask. Although we do not submit an official submission to the sentence-level subtask, we show that this subword-based approach is highly effective there as well.

Publication details

Authors in the community:

Publication version

VoR - Version of Record

Publisher

Association for Computational Linguistics

Title of the publication container

19th SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology

First page or article number

131

Last page

138

Fields of Science and Technology (FOS)

computer-and-information-sciences - Computer and information sciences

Publication language (ISO code)

eng - English

Rights type:

Open access

Creative Commons license

CC-BY - Attribution