Unsupervised Pidgin Text Generation By Pivoting English Data and Self-Training

Abstract

West African Pidgin English is a language that is significantly spoken in West Africa, consisting of at least 75 million speakers. Nevertheless, proper machine translation systems and relevant NLP datasets for pidgin English are virtually absent. In this work, we develop techniques targeted at bridging the gap between Pidgin English and English in the context of natural language generation. %As a proof of concept, we explore the proposed techniques in the area of data-to-text generation. By building upon the previously released monolingual Pidgin English text and parallel English data-to-text corpus, we hope to build a system that can automatically generate Pidgin English descriptions from structured data. We first train a data-to-English text generation system, before employing techniques in unsupervised neural machine translation and self-training to establish the Pidgin-to-English cross-lingual alignment. The human evaluation performed on the generated Pidgin texts shows that, though still far from being practically usable, the pivoting + self-training technique improves both Pidgin text fluency and relevance.

arXiv commentAccepted to Workshop at ICLR 2020

References in corpus (6)

• Unsupervised Machine Translation Using Monolingual Corpora Only
• Revisiting Self-Training for Neural Sequence Generation
• Unsupervised Neural Machine Translation
• Zero-Shot Paraphrase Generation with Multilingual Language Models
• PidginUNMT: Unsupervised Neural Machine Translation from West African Pidgin to English
• Neobility at SemEval-2017 Task 1: An Attention-based Sentence Similarity Model

Cited by in corpus (8)

• Neural Data-to-Text Generation via Jointly Learning the Segmentation and Correspondence
• Safe Handover in Mixed-Initiative Control for Cyber-Physical Systems
• Diversifying Dialogue Generation with Non-Conversational Text
• DART: A Lightweight Quality-Suggestive Data-to-Text Annotation Tool
• Learning Fine-grained Fact-Article Correspondence in Legal Cases
• Neural Data-to-Text Generation with LM-based Text Augmentation
• Jointly Improving Language Understanding and Generation with Quality-Weighted Weak Supervision of Automatic Labeling
• Does the Order of Training Samples Matter? Improving Neural Data-to-Text Generation with Curriculum Learning