SynthBio: A Case Study in Faster Curation of Text Datasets

Ann Yuan, Daphne Ippolito, Vitaly Nikolaev, Chris Callison-Burch, Andy Coenen, Sebastian Gehrmann

Keywords: NLP, machine learning, datasets, NLG

TL;DR: We introduce a dataset curation method in which a language model generates seed text which human raters edit - and use the method to curate SynthBio - a new structure-to-text dataset of biographies describing fictional individuals.

Abstract: NLP researchers need more, higher-quality text datasets. Human-labeled datasets are expensive to collect, while datasets collected via automatic retrieval from the web such as WikiBio [Lebret 2016] are noisy and can include undesired biases. Moreover, data sourced from the web is often included in datasets used to pretrain models, leading to inadvertent cross-contamination of training and test sets. In this work we introduce a novel method for efficient dataset curation: we use a large language model to provide seed generations to human raters, thereby changing dataset authoring from a writing task to an editing task. We use our method to curate SynthBio - a new evaluation set for WikiBio - comprised of structured attribute lists describing fictional individuals, mapped to natural language biographies. We show that our dataset of fictional biographies is less noisy than WikiBio, and also more balanced with respect to gender and nationality.

URL: https://storage.googleapis.com/gem-benchmark/SynthBio.json

Contribution Process Agreement: Yes

Dataset Url: https://storage.googleapis.com/gem-benchmark/SynthBio.json

Author Statement: Yes