Controllable Text Generation with Language Constraints
This work addresses the challenge of controllable text generation for AI applications, though it is incremental as it builds on existing methods with a new benchmark and guidance approach.
The authors tackled the problem of text generation with natural language constraints by creating a benchmark Cognac and proposing CognacGen, a method that uses a language model's internal knowledge to guide generation, resulting in improved performance over baselines on constraint-conforming text.
We consider the task of text generation in language models with constraints specified in natural language. To this end, we first create a challenging benchmark Cognac that provides as input to the model a topic with example text, along with a constraint on text to be avoided. Unlike prior work, our benchmark contains knowledge-intensive constraints sourced from databases like Wordnet and Wikidata, which allows for straightforward evaluation while striking a balance between broad attribute-level and narrow lexical-level controls. We find that even state-of-the-art language models like GPT-3 fail often on this task, and propose a solution to leverage a language model's own internal knowledge to guide generation. Our method, called CognacGen, first queries the language model to generate guidance terms for a specified topic or constraint, and uses the guidance to modify the model's token generation probabilities. We propose three forms of guidance (binary verifier, top-k tokens, textual example), and employ prefix-tuning approaches to distill the guidance to tackle diverse natural language constraints. Through extensive empirical evaluations, we demonstrate that CognacGen can successfully generalize to unseen instructions and outperform competitive baselines in generating constraint conforming text.