Pre-training via Paraphrasing
Published in
Neural Information Processing Systems(2020)External Links:
TL
TL;DR
It is shown that fine-tuning gives strong performance on a range of discriminative and generative tasks in many languages, making MARGE the most generally applicable pre-training method to date.
Abstract
We introduce MARGE, a pre-trained sequence-to-sequence model learned with an unsupervised multi-lingual multi-document paraphrasing objective. MARGE provides an alternative to the dominant masked language modeling paradigm, where we self-supervise the reconstruction of target text by retrieving a set of related texts (in many languages) and conditioning on them to maximize the likelihood of generating the original. We show it is possible to jointly learn to do retrieval and reconstruction, given only a random initialization. The objective noisily captures aspects of paraphrase, translation, multi-document summarization, and information retrieval, allowing for strong zero-shot performance on several tasks. For example, with no additional task-specific training we achieve BLEU scores of up to 35.8 for document translation. We further show that fine-tuning gives strong performance on a range of discriminative and generative tasks in many languages, making MARGE the most generally applicable pre-training method to date.
References38 items
1
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
2
MLSUM: The Multilingual Summarization Corpus
3
XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization
4
A Primer in BERTology: What We Know About How BERT Works
5