Takeishi Yuka

Takeishi Yuka

March 31, 2026

Waka, Poetry generation, Artificial intelligence(AI), Deep learning.

This paper employs a deep learning-based Waka generation model to computer-generate AI waka. From the perspective of AI waka borrowing human-created waka phrases, this paper compares and analyzes the quality of AI waka generated by the WakaVT model and two benchmark models. The analysis results indicate that, relative to the benchmark models, the latent variable sequence enhances the diversity of AI waka generated by the WakaVT model, effectively avoiding the issue of excessive borrowing of human-created Waka phrases in generated AI waka. The WakaVT model has successfully learned the vocabulary, grammar, and logical structure between sentences of human-created waka, enabling it to generate creative AI waka through novel word combinations and metaphorical techniques.

© 2026, the Authors. Published by ISCCAC

This is an open access article distributed under the CC BY-NC license