Publisher's Synopsis
Hybrid text production today comprises machine translation chunks, human-produced paragraphs modelled on the writer's L1, and generative AI contributions via prompting. This book explores how corpus-based cross-linguistic studies can contribute to building and optimising Natural Language Generation in multilingual contexts. The aim is to show how rich linguistic annotation can provide valuable data to enhance and refine hybrid text production. Corpus data are mined from CLANES, a multilayer English-Spanish comparable corpus annotated for PoS, semantic, rhetorical and pragmatic information. The study highlights the implications of leveraging LLMs for the automatic generation of domain-specific texts. It also addresses the new challenges and opportunities raised by AI-enhanced data augmentation and post-editing, stressing the need for human control over text generation.
