Screening through a broad pool: Towards better diversity for lexically constrained text generation

• Main Authors: YUAN, Changsen, HUANG, Heyan, CAO, Yixin, CAO, Qianwen
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2024
• Subjects: Constrained text generation; Pre-trained language models; Randomly insert; Randomly mask; Text diversity; Databases and Information Systems; Theory and Algorithms
• Online Access: https://ink.library.smu.edu.sg/sis_research/8478; https://ink.library.smu.edu.sg/context/sis_research/article/9481/viewcontent/ScreeningBroadPool_av.pdf
• Institution: Singapore Management University

Lexically constrained text generation (CTG) is to generate text that contains given constrained keywords. However, the text diversity of existing models is still unsatisfactory. In this paper, we propose a lightweight dynamic refinement strategy that aims at increasing the randomness of inference to...