PIP: Parse-Instructed Prefix for Syntactically Controlled Paraphrase Generation

TL;DR

This paper introduces PIP, a parameter-efficient prefix-tuning method for syntactically controlled paraphrase generation that reduces training costs and improves syntax control accuracy over traditional fine-tuning methods.

Contribution

Proposes PIP, a novel prefix-tuning approach that captures syntax knowledge efficiently, requiring fewer parameters and less training data for paraphrase generation.

Findings

PIP achieves higher syntax control accuracy than traditional fine-tuning.

PIP uses 10 times fewer learnable parameters.

PIP outperforms existing prefix-tuning methods in syntax capture.

Abstract

Syntactically controlled paraphrase generation requires language models to generate paraphrases for sentences according to specific syntactic structures. Existing fine-tuning methods for this task are costly as all the parameters of the model need to be updated during the training process. Inspired by recent studies on parameter-efficient learning, we propose Parse-Instructed Prefix (PIP), a novel adaptation of prefix-tuning to tune large pre-trained language models on syntactically controlled paraphrase generation task in a low-data setting with significantly less training cost. We introduce two methods to instruct a model's encoder prefix to capture syntax-related knowledge: direct initiation (PIP-Direct) and indirect optimization (PIP-Indirect). In contrast to traditional fine-tuning methods for this task, PIP is a compute-efficient alternative with 10 times less learnable parameters. Compared to existing prefix-tuning methods, PIP excels at capturing syntax control information, achieving significantly higher performance at the same level of learnable parameter count.