2021/01/28 by Hrituraj Singh, Singh, Hrituraj, Gaurav Verma +5
Computer Science · #Artificial Intelligence (cs.AI) #Authorship Attribution and Profiling #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.2101.11836
openalex publication_date 2021/01/28 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Author stylized rewriting is the task of rewriting an input text in a\nparticular author's style. Recent works in this area have leveraged\nTransformer-based language models in a denoising autoencoder setup to generate\nauthor stylized text without relying on a parallel corpus of data. However,\nthese approaches are limited by the lack of explicit control of target\nattributes and being entirely data-driven. In this paper, we propose a\nDirector-Generator framework to rewrite content in the target author's style,\nspecifically focusing on certain target attributes. We show that our proposed\nframework works well even with a limited-sized target author corpus. Our\nexperiments on corpora consisting of relatively small-sized text authored by\nthree distinct authors show significant improvements upon existing works to\nrewrite input texts in target author's style. Our quantitative and qualitative\nanalyses further show that our model has better meaning retention and results\nin more fluent generations.\n