Unsupervised Controllable Text Formalization

TL;DR

This paper introduces an unsupervised neural framework for controllable text formalization, enabling style adjustment without parallel data, and demonstrates its effectiveness in transforming text to more formal styles with user-controlled degrees of formality.

Contribution

The paper presents a novel unsupervised training scheme for controllable text transformation using auxiliary scorers, eliminating the need for parallel corpora.

Findings

Effective formalization of input text with controllable degree of formality

Successful transformation to more formal style on public datasets

Code and datasets released for academic use

Abstract

We propose a novel framework for controllable natural language transformation. Realizing that the requirement of parallel corpus is practically unsustainable for controllable generation tasks, an unsupervised training scheme is introduced. The crux of the framework is a deep neural encoder-decoder that is reinforced with text-transformation knowledge through auxiliary modules (called scorers). The scorers, based on off-the-shelf language processing tools, decide the learning scheme of the encoder-decoder based on its actions. We apply this framework for the text-transformation task of formalizing an input text by improving its readability grade; the degree of required formalization can be controlled by the user at run-time. Experiments on public datasets demonstrate the efficacy of our model towards: (a) transforming a given text to a more formal style, and (b) introducing appropriate amount of formalness in the output text pertaining to the input control. Our code and datasets are released for academic use.