LAD: Language Models as Data for Zero-Shot Dialog

Shikib Mehri, Yasemin Altun, Maxine Eskenazi
7/28/2022
cs.CLcs.AI

Official code: Shikib/lad (8 stars, last commit 2022-07-23).

Abstract

To facilitate zero-shot generalization in taskoriented dialog, this paper proposes Language Models as Data (LAD). LAD is a paradigm for creating diverse and accurate synthetic data which conveys the necessary structural constraints and can be used to train a downstream neural dialog model. LAD leverages GPT-3 to induce linguistic diversity. LAD achieves significant performance gains in zero-shot settings on intent prediction (+15%), slot filling (+31.4 F-1) and next action prediction (+11 F1). Furthermore, an interactive human evaluation shows that training with LAD is competitive with training on human dialogs. LAD is open-sourced, with the code and data available at https://github.com/Shikib/lad.

View on arXivView PDF

Code Implementations(1)

Shikib/ladOfficial100%

Code for LAD: Language Models as Data for Zero-Shot Dialog (SIGdial 2022)

80Jul 23, 20224 years ago

Similar Papers

Few-shot Natural Language Generation for Task-Oriented Dialog

88% match1 repo1902020

Zero-Shot Dialog Generation with Cross-Domain Latent Actions

87% match2 repos1322018

InstructTODS: Large Language Models for End-to-End Task-Oriented Dialogue Systems

84% match1 repo182023

Instances Need More Care: Rewriting Prompts for Instances with LLMs in the Loop Yields Better Zero-Shot Performance

84% match1 repo32023

I$^2$KD-SLU: An Intra-Inter Knowledge Distillation Framework for Zero-Shot Cross-Lingual Spoken Language Understanding

84% match2023

A Preliminary Evaluation of ChatGPT for Zero-shot Dialogue Understanding

83% match2023

Cite this paper

@article{mehri2022lad,
  title  = {LAD: Language Models as Data for Zero-Shot Dialog},
  author = {Shikib Mehri and Yasemin Altun and Maxine Eskenazi},
  year   = {2022},
  eprint = {2207.14393},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2207.14393v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2207.14393.svg)](https://codeofpaper.com/paper/2207.14393)

Discussion