LiFi: Lightweight Controlled Text Generation with Fine-Grained Control Codes
This work addresses the need for more precise control mechanisms in text generation for applications requiring nuanced outputs, representing an incremental advancement over previous methods that used discrete control codes.
The paper tackled the problem of achieving fine-grained control in text generation by introducing LIFI, a lightweight method that uses continuous, relative, and nonexclusive control codes derived from an attribute classifier, resulting in substantial performance improvements over existing baselines on tasks like sentiment control, topic control, and stylistic novel writing.
In the rapidly evolving field of text generation, the demand for more precise control mechanisms has become increasingly apparent. To address this need, we present a novel methodology, LIFI, which offers a lightweight approach with fine-grained control for controlled text generation. Unlike previous studies that train pre-trained language models to follow discrete, categorical, and exclusive control codes, LIFI learns controlled text generation under the guidance of continuous, relative, and nonexclusive control codes. These fine-grained codes are automatically derived from an attribute classifier, initially trained with a small amount of labeled data and subsequently employed to label abundant unlabeled data, thus garnering more extensive supervision signals. Moreover, to achieve efficient control, we incorporate the fine-grained control codes with adapters, a parameter- and compute-efficient way to steer a pre-trained language model. We evaluate LIFI on two conventional tasks -- sentiment control and topic control -- and one newly proposed task -- stylistic novel writing. Comprehensive experimental results validate the effectiveness of our proposed methods, demonstrating substantial performance improvements over existing baselines.