Transforming Multi-Conditioned Generation from Meaning Representation

TL;DR

This paper introduces a one-stage GPT2-based framework for multi-conditioned language generation from meaning representations, simplifying the process and achieving competitive results with less data and zero-shot capabilities.

Contribution

Proposes a novel single-step generation model directly from meaning representations, eliminating the need for sentence planning and surface realization stages.

Findings

Comparable performance to previous systems on E2E dataset

Effective zero-shot generation with only 10% of training data

Demonstrates the simplicity and efficiency of the proposed approach

Abstract

In task-oriented conversation systems, natural language generation systems that generate sentences with specific information related to conversation flow are useful. Our study focuses on language generation by considering various information representing the meaning of utterances as multiple conditions of generation. NLG from meaning representations, the conditions for sentence meaning, generally goes through two steps: sentence planning and surface realization. However, we propose a simple one-stage framework to generate utterances directly from MR (Meaning Representation). Our model is based on GPT2 and generates utterances with flat conditions on slot and value pairs, which does not need to determine the structure of the sentence. We evaluate several systems in the E2E dataset with 6 automatic metrics. Our system is a simple method, but it demonstrates comparable performance to previous systems in automated metrics. In addition, using only 10\% of the data set without any other techniques, our model achieves comparable performance, and shows the possibility of performing zero-shot generation and expanding to other datasets.