Formality Style Transfer for Noisy, User-generated Conversations: Extracting Labeled, Parallel Data from Unlabeled Corpora

Isak Czeresnia Etinger, Alan W Black


Abstract
Typical datasets used for style transfer in NLP contain aligned pairs of two opposite extremes of a style. As each existing dataset is sourced from a specific domain and context, most use cases will have a sizable mismatch from the vocabulary and sentence structures of any dataset available. This reduces the performance of the style transfer, and is particularly significant for noisy, user-generated text. To solve this problem, we show a technique to derive a dataset of aligned pairs (style-agnostic vs stylistic sentences) from an unlabeled corpus by using an auxiliary dataset, allowing for in-domain training. We test the technique with the Yahoo Formality Dataset and 6 novel datasets we produced, which consist of scripts from 5 popular TV-shows (Friends, Futurama, Seinfeld, Southpark, Stargate SG-1) and the Slate Star Codex online forum. We gather 1080 human evaluations, which show that our method produces a sizable change in formality while maintaining fluency and context; and that it considerably outperforms OpenNMT’s Seq2Seq model directly trained on the Yahoo Formality Dataset. Additionally, we publish the full pipeline code and our novel datasets.
Anthology ID:
D19-5502
Volume:
Proceedings of the 5th Workshop on Noisy User-generated Text (W-NUT 2019)
Month:
November
Year:
2019
Address:
Hong Kong, China
Venues:
EMNLP | WNUT | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
11–16
Language:
URL:
https://www.aclweb.org/anthology/D19-5502
DOI:
10.18653/v1/D19-5502
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
http://aclanthology.lst.uni-saarland.de/D19-5502.pdf
Attachment:
 D19-5502.Attachment.zip