Unwritten languages demand attention too! Word discovery with encoder-decoder models

Word discovery is the task of extracting words from un-segmented text. In this paper we examine to what extent neural networks can be applied to this task in a realistic unwritten language scenario, where only small corpora and limited annotations are available. We investigate two scenarios: one with no supervision and another with limited supervision with access to the most frequent words. Obtained results show that it is possible to retrieve at least 27% of the gold standard vocabulary by training an encoder-decoder neural machine translation system with only 5,157 sentences. This result is close to those obtained with a task-specific Bayesian nonparametric model. Moreover, our approach has the advantage of generating translation alignments, which could be used to create a bilingual lexicon. As a future perspective, this approach is also well suited to work directly from speech.

• Publication year

  2017

• Venue

  Automatic Speech Recognition & Understanding

• Publication date

  2017-09-17

• Fields of study

  Linguistics, Computer Science

• Identifiers
  DOI 10.1109/ASRU.2017.8268972arXiv 1709.05631
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• A case study on using speech-to-translation alignments for language documentation

  2017cited by this paper
• Sequence-to-Sequence Models Can Directly Transcribe Foreign Speech

  2017cited by this paper
• Six Challenges for Neural Machine Translation

  2017cited by this paper
• Sequence-to-Sequence Models Can Directly Translate Foreign Speech

  2017cited by this paper
• Breaking the Unwritten Language Barrier: The BULB Project

  2016cited by this paper
• Weakly supervised spoken term discovery using cross-lingual side information

  2016cited by this paper
• The Zero Resource Speech Challenge 2015: Proposed Approaches and Results

  2016cited by this paper
• Learning a Lexicon and Translation Model from Phoneme Lattices

  2016cited by this paper
• Toward human-assisted lexical unit discovery without text resources

  2016cited by this paper
• Fully Character-Level Neural Machine Translation without Explicit Segmentation

  2016cited by this paper
• An Attentional Model for Speech Translation Without Transcription

  2016influential reference
• Parallel Speech Collection for Under-resourced Language Studies Using the Lig-Aikuma Mobile Device App

  2016cited by this paper
• Morphological Segmentation with Window LSTM Neural Networks

  2016cited by this paper
• Preliminary Experiments on Unsupervised Word Discovery in Mboshi

  2016influential reference
• Phoneme Boundary Detection using Deep Bidirectional LSTMs

  2016cited by this paper
• Translate : A Proof of Concept for End-to-End Speech-to-Text Translation

  2016influential reference
• Unsupervised Lexicon Discovery from Acoustic Input

  2015cited by this paper
• Inducing Bilingual Lexicons from Small Quantities of Sentence-Aligned Phonemic Transcriptions

  2015cited by this paper
• Neural Machine Translation by Jointly Learning to Align and Translate

  2014influential reference
• A Joint Learning Model of Word Segmentation, Lexical Acquisition, and Phonetic Variability

  2013cited by this paper
• A summary of the 2012 JHU CLSP workshop on zero resource speech technologies and models of early language acquisition

  2013cited by this paper
• The Cambridge handbook of endangered languages

  2011cited by this paper
• Recession Segmentation: Simpler Online Word Segmentation Using Limited Resources

  2010cited by this paper
• A Bayesian framework for word segmentation: exploring the effects of context.

  2009influential reference
• Improving nonparameteric Bayesian inference: experiments on unsupervised word segmentation with adaptor grammars

  2009cited by this paper
• Nonparametric bayesian models of lexical acquisition

  2007cited by this paper
• Moses: Open Source Toolkit for Statistical Machine Translation

  2007cited by this paper
• TOWARDS SPEECH TRANSLATION OF NON WRITTEN LANGUAGES

  2006cited by this paper
• Contextual Dependencies in Unsupervised Word Segmentation

  2006cited by this paper
• Applications and Explanations of Zipf’s Law

  1998cited by this paper
• Proceedings of the Joint Conferences on New Methods in Language Processing and Computational Natural Language Learning

  1998cited by this paper

• Learning Through Transcription

  2022cites this paper
• Unsupervised Word Segmentation from Discrete Speech Units in Low-Resource Settings

  2021cites this paper
• Sparse Transcription

  2021cites this paper
• AlloST: Low-resource Speech Translation without Source Transcription

  2021cites this paper
• Local Word Discovery for Interactive Transcription

  2021cites this paper
• On the Difficulty of Segmenting Words with Attention

  2021cites this paper
• AlloVera: A Multilingual Allophone Database

  2020cites this paper
• Investigating Language Impact in Bilingual Approaches for Computational Language Documentation

  2020cites this paper
• ORTHROS: non-autoregressive end-to-end speech translation With dual-decoder

  2020cites this paper
• Investigating alignment interpretability for low-resource NMT

  2020cites this paper
• Unsupervised word discovery for computational language documentation

  2019cites this paper
• Multilingual End-to-End Speech Translation

  2019cites this paper
• C L ] 1 O ct 2 01 9 MULTILINGUAL END-TO-END SPEECH TRANSLATION

  2019cites this paper
• Controlling Utterance Length in NMT-based Word Segmentation with Attention

  2019influential citation
• Weakly Supervised Attention Networks for Entity Recognition

  2019cites this paper
• Tied Multitask Learning for Neural Speech Translation

  2018influential citation
• Collecter, Transcrire, Analyser : quand la machine assiste le linguiste dans son travail de terrain. (Collecting, Transcribing, Analyzing : Machine-Assisted Linguistic Fieldwork)

  2018cites this paper
• A small Griko-Italian speech translation corpus

  2018cites this paper
• Unsupervised Word Segmentation from Speech with Attention

  2018influential citation
• Linguistic Unit Discovery from Multi-Modal Inputs in Unwritten Languages: Summary of the “Speaking Rosetta” JSALT 2017 Workshop

  2018cites this paper
• A Very Low Resource Language Speech Corpus for Computational Language Documentation Experiments

  2017cites this paper