Multi-Style Transfer with Discriminative Feedback on Disjoint Corpus

Abstract

Style transfer has been widely explored in natural language generation with non-parallel corpus by directly or indirectly extracting a notion of style from source and target domain corpus. A common shortcoming of existing approaches is the prerequisite of joint annotations across all the stylistic dimensions under consideration. Availability of such dataset across a combination of styles limits the extension of these setups to multiple style dimensions. While cascading single-dimensional models across multiple styles is a possibility, it suffers from content loss, especially when the style dimensions are not completely independent of each other. In our work, we relax this requirement of jointly annotated data across multiple styles by using independently acquired data across different style dimensions without any additional annotations. We initialize an encoder-decoder setup with transformerbased language model pre-trained on a generic corpus and enhance its re-writing capability to multiple target style dimensions by employing multiple style-aware language models as discriminators. Through quantitative and qualitative evaluation, we show the ability of our model to control styles across multiple style dimensions while preserving content of the input text. We compare it against baselines involving cascaded state-of-the-art uni-dimensional style transfer models.

Are You Robert or RoBERTa? Deceiving Online Authorship Attribution Models Using Neural Text Generators

“…The authors also proposed novel metrics for measuring the degree of authorial stylistic alignment, leveraging lexical and syntactical patterns to score similarities in writing styles. Moreover, in (Goyal et al 2021), the authors combined multiple 'style aware' natural language models to allow for the transference of both sentiment and formality. This approach was able to outperform that of Syed et al (2020) in terms of style transference, though not in terms of content preservationachieving strong performances in both of these aspects remains an open challenge.…”

Entailment Relation Aware Paraphrase Generation

“…sentiment, formality). RL has been used to explicitly reward the output to adhere to a target attribute (Gong et al 2019;Sancheti et al 2020;Luo et al 2019;Liu, Neubig, and Wieting 2020;Goyal et al 2021). The target attributes are only a function of the output and defined at a lexical level.…”

Are You Robert or RoBERTa? Deceiving Online Authorship Attribution Models Using Neural Text Generators

“…The authors also proposed novel metrics for measuring the degree of authorial stylistic alignment, leveraging lexical and syntactical patterns to score similarities in writing styles. Moreover, in (Goyal et al 2021), the authors combined multiple 'style aware' natural language models to allow for the transferal of both sentiment and formality. This approach was able to outperform that of Syed et al (2020) in terms of style transference, though not in terms of content preservation, and achieving strong performs in both of these aspects remains a challenge.…”