Zero-shot Sonnet Generation with Discourse-level Planning and Aesthetics Features
Poetry generation, and creative language generation in general, usually suffers from the lack of large training data. In this paper, we present a novel framework to generate sonnets that does not require training on poems. We design a hierarchical framework which plans the poem sketch before decoding. Specifically, a content planning module is trained on non-poetic texts to obtain discourse-level coherence; then a rhyme module generates rhyme words and a polishing module introduces imagery and similes for aesthetics purposes. Finally, we design a constrained decoding algorithm to impose the meter-and-rhyme constraint of the generated sonnets. Automatic and human evaluation show that our multi-stage approach without training on poem corpora generates more coherent, poetic, and creative sonnets than several strong baselines.
Code (1)
Tasks
Sonnet GenerationText GenerationSimilar Papers 제목 키워드 기반
Sonnet Generation by Training on Non-poetic Texts with Discourse-level Coherence and Poetic Features
Poetry generation, and creative language generation in general, usually suffers from the lack of large training data. In this paper, we present a novel framework to generate sonnets that does not require training on poem…
Sonnet GenerationText GenerationZero-shot Chinese Discourse Dependency Parsing via Cross-lingual Mapping
Due to the absence of labeled data, discourse parsing still remains challenging in some languages. In this paper, we present a simple and efficient method to conduct zero-shot Chinese text-level dependency parsing by lev…
Dependency ParsingDiscourse ParsingSentenceLong Live Fine-Tuning: Task-Specific Transformers Outperform Zero-Shot LLMs for Misinformation Response Classification on Reddit
As large language models (LLMs) become default tools for online information verification, an implicit assumption follows them: that scale and general capability are sufficient for nuanced classification of misinformation…
In-Context Learning for Long-Context Sentiment Analysis on Infrastructure Project Opinions
Large language models (LLMs) have achieved impressive results across various tasks. However, they still struggle with long-context documents. This study evaluates the performance of three leading LLMs: GPT-4o, Claude 3.5…
In-Context LearningSentiment AnalysisAnalysis of LLM Performance on AWS Bedrock: Receipt-item Categorisation Case Study
This paper presents a systematic, cost-aware evaluation of large language models (LLMs) for receipt-item categorisation within a production-oriented classification framework. We compare four instruction-tuned models avai…