Learning Text Styles: A Study on Transfer, Attribution, and Verification
It solves problems in computational linguistics for researchers and practitioners by advancing text style analysis, though it appears incremental in its approach.
This thesis tackles the problem of understanding and manipulating text styles by addressing challenges in text style transfer, authorship attribution, and authorship verification, using methods like parameter-efficient LLM adaptation and contrastive disentanglement to achieve improvements in these tasks.
This thesis advances the computational understanding and manipulation of text styles through three interconnected pillars: (1) Text Style Transfer (TST), which alters stylistic properties (e.g., sentiment, formality) while preserving content; (2)Authorship Attribution (AA), identifying the author of a text via stylistic fingerprints; and (3) Authorship Verification (AV), determining whether two texts share the same authorship. We address critical challenges in these areas by leveraging parameter-efficient adaptation of large language models (LLMs), contrastive disentanglement of stylistic features, and instruction-based fine-tuning for explainable verification.