LOGEN: Few-shot Logical Knowledge-Conditioned Text Generation with Self-training
This addresses the challenge of data-hungry methods for logical knowledge-conditioned text generation, making it more applicable to real-world scenarios with limited data.
The paper tackles the problem of generating text from structured data with logical knowledge in few-shot settings, achieving better performance than baselines using only 20 to 100 seed logical forms through self-training and pseudo-form sampling.
Natural language generation from structured data mainly focuses on surface-level descriptions, suffering from uncontrollable content selection and low fidelity. Previous works leverage logical forms to facilitate logical knowledge-conditioned text generation. Though achieving remarkable progress, they are data-hungry, which makes the adoption for real-world applications challenging with limited data. To this end, this paper proposes a unified framework for logical knowledge-conditioned text generation in the few-shot setting. With only a few seeds logical forms (e.g., 20/100 shot), our approach leverages self-training and samples pseudo logical forms based on content and structure consistency. Experimental results demonstrate that our approach can obtain better few-shot performance than baselines.