Controlling Linguistic Style Aspects in Neural Language Generation

TL;DR

This paper explores controlling stylistic aspects in neural language generation by conditioning RNN models on content and style parameters, demonstrated on movie reviews to produce coherent, stylistically aligned text.

Contribution

It introduces a conditioned RNN approach for controlling both content and stylistic features in neural language generation, extending beyond content-focused models.

Findings

Successful control of stylistic aspects in generated text

Generation of coherent sentences matching style and content

Effective application in movie review domain

Abstract

Most work on neural natural language generation (NNLG) focus on controlling the content of the generated text. We experiment with controlling several stylistic aspects of the generated text, in addition to its content. The method is based on conditioned RNN language model, where the desired content as well as the stylistic parameters serve as conditioning contexts. We demonstrate the approach on the movie reviews domain and show that it is successful in generating coherent sentences corresponding to the required linguistic style and content.