Few-Shot Dialogue Summarization via Skeleton-Assisted Prompt Transfer in Prompt Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Kaige, Yu, Tong, Wang, Haoliang, Wu, Junda, Zhao, Handong, Zhang, Ruiyi, Mahadik, Kanak, Nenkova, Ani, Riedl, Mark
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929257163259904
author Xie, Kaige
Yu, Tong
Wang, Haoliang
Wu, Junda
Zhao, Handong
Zhang, Ruiyi
Mahadik, Kanak
Nenkova, Ani
Riedl, Mark
author_facet Xie, Kaige
Yu, Tong
Wang, Haoliang
Wu, Junda
Zhao, Handong
Zhang, Ruiyi
Mahadik, Kanak
Nenkova, Ani
Riedl, Mark
contents In real-world scenarios, labeled samples for dialogue summarization are usually limited (i.e., few-shot) due to high annotation costs for high-quality dialogue summaries. To efficiently learn from few-shot samples, previous works have utilized massive annotated data from other downstream tasks and then performed prompt transfer in prompt tuning so as to enable cross-task knowledge transfer. However, existing general-purpose prompt transfer techniques lack consideration for dialogue-specific information. In this paper, we focus on improving the prompt transfer from dialogue state tracking to dialogue summarization and propose Skeleton-Assisted Prompt Transfer (SAPT), which leverages skeleton generation as extra supervision that functions as a medium connecting the distinct source and target task and resulting in the model's better consumption of dialogue state information. To automatically extract dialogue skeletons as supervised training data for skeleton generation, we design a novel approach with perturbation-based probes requiring neither annotation effort nor domain knowledge. Training the model on such skeletons can also help preserve model capability during prompt transfer. Our method significantly outperforms existing baselines. In-depth analyses demonstrate the effectiveness of our method in facilitating cross-task knowledge transfer in few-shot dialogue summarization.
format Preprint
id arxiv_https___arxiv_org_abs_2305_12077
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Few-Shot Dialogue Summarization via Skeleton-Assisted Prompt Transfer in Prompt Tuning
Xie, Kaige
Yu, Tong
Wang, Haoliang
Wu, Junda
Zhao, Handong
Zhang, Ruiyi
Mahadik, Kanak
Nenkova, Ani
Riedl, Mark
Computation and Language
In real-world scenarios, labeled samples for dialogue summarization are usually limited (i.e., few-shot) due to high annotation costs for high-quality dialogue summaries. To efficiently learn from few-shot samples, previous works have utilized massive annotated data from other downstream tasks and then performed prompt transfer in prompt tuning so as to enable cross-task knowledge transfer. However, existing general-purpose prompt transfer techniques lack consideration for dialogue-specific information. In this paper, we focus on improving the prompt transfer from dialogue state tracking to dialogue summarization and propose Skeleton-Assisted Prompt Transfer (SAPT), which leverages skeleton generation as extra supervision that functions as a medium connecting the distinct source and target task and resulting in the model's better consumption of dialogue state information. To automatically extract dialogue skeletons as supervised training data for skeleton generation, we design a novel approach with perturbation-based probes requiring neither annotation effort nor domain knowledge. Training the model on such skeletons can also help preserve model capability during prompt transfer. Our method significantly outperforms existing baselines. In-depth analyses demonstrate the effectiveness of our method in facilitating cross-task knowledge transfer in few-shot dialogue summarization.
title Few-Shot Dialogue Summarization via Skeleton-Assisted Prompt Transfer in Prompt Tuning
topic Computation and Language
url https://arxiv.org/abs/2305.12077