A Plug-and-Play Method for Controlled Text Generation
This addresses the need for simple, effective control in text generation tasks like story generation, offering a plug-and-play solution without requiring additional models or fine-tuning, though it is incremental as it builds on existing decoding methods.
The paper tackles the problem of controlling text generation from large pre-trained language models to ensure specific words are included, presenting a plug-and-play decoding method that adds a shift to the probability distribution towards semantically similar words, which outperforms competing methods in human evaluations and guarantees guide word appearance without impacting fluency.
Large pre-trained language models have repeatedly shown their ability to produce fluent text. Yet even when starting from a prompt, generation can continue in many plausible directions. Current decoding methods with the goal of controlling generation, e.g., to ensure specific words are included, either require additional models or fine-tuning, or work poorly when the task at hand is semantically unconstrained, e.g., story generation. In this work, we present a plug-and-play decoding method for controlled language generation that is so simple and intuitive, it can be described in a single sentence: given a topic or keyword, we add a shift to the probability distribution over our vocabulary towards semantically similar words. We show how annealing this distribution can be used to impose hard constraints on language generation, something no other plug-and-play method is currently able to do with SOTA language generators. Despite the simplicity of this approach, we see it works incredibly well in practice: decoding from GPT-2 leads to diverse and fluent sentences while guaranteeing the appearance of given guide words. We perform two user studies, revealing that (1) our method outperforms competing methods in human evaluations; and (2) forcing the guide words to appear in the generated text has no impact on the fluency of the generated text.