Arabic poetry generation controlled by meaning meter and length
Shaer: Controlled Arabic Poetry Generation with Meter Subform and Semantic Conditioning
Computation and Language
Summary
Generating classical Arabic poetry is tricky because it must follow strict rules about meaning, language, rhythm, and length. The authors created Shaer, a system that can generate poems by carefully controlling the poem’s theme, the specific rhythm pattern, and the number of lines. They built a large dataset of Arabic poems labeled with these details and trained a specialized language model that performs well in meeting all these constraints. Tests show Shaer creates poems that closely follow the poetic rules and fit requested themes better than previous models.
What this means in practice
- •For natural language generation developers: Generate Arabic poems that strictly follow meter and semantic themes for creative writing tools or digital assistants.
- •For language learning tool builders: Create educational content that demonstrates correct classical Arabic poetic forms with matching semantic descriptions and lengths.
Authors
Ahmad Abbas, Tamara Fakih, Nour Fakih, Ammar Mohanna
Abstract
Classical Arabic poetry generation requires simultaneously satisfying semantic, linguistic, and fine-grained prosodic constraints. Existing systems typically control broad poetic attributes but do not jointly model semantic intent, meter subform, and poem length. We present Shaer, a controllable Classical Arabic poetry generation framework jointly conditioned on natural-language descriptions, meter subforms, and target hemistich counts. To support this task, we construct an enriched corpus of 116,032 classical Arabic poems derived from Ashaar, containing normalized meter-subform labels and automatically generated, validated semantic descriptions. We then adapt Yehia-7B using QLoRA-based supervised fine-tuning with a completion-only objective. Our evaluation combines automatic assessment of base-meter conformity, requested-subform adherence, and length control with three LLM judges, blinded human evaluation, and memorization analysis. Shaer achieves 95.17% base-meter accuracy, 91.75% poem-level meter-subform accuracy, and 83.40% exact count accuracy. Relative to its untuned foundation model, these results represent gains of 68.68, 57.77, and 38.93 percentage points, respectively; Shaer also attains the highest base-meter accuracy among all evaluated systems. Multi-LLM evaluation and a blinded human assessment of top-ranked outputs further indicate competitive semantic and literary quality. Finally, analysis of all 3,481 test generations finds no exact copies from the training corpus or paired source poems. Code, models, and datasets are publicly available.