Neural data-to-text generation: A comparison between pipeline and end-to-end architectures

TL;DR

This paper compares traditional pipeline and modern end-to-end neural approaches for data-to-text generation from RDF triples, finding that pipeline models produce higher quality and better generalization.

Contribution

It provides a systematic comparison of pipeline and end-to-end neural architectures using state-of-the-art methods, with comprehensive evaluations.

Findings

Pipeline models generate more accurate and coherent texts.

Pipeline approaches generalize better to unseen data.

End-to-end models are less effective in maintaining quality.

Abstract

Traditionally, most data-to-text applications have been designed using a modular pipeline architecture, in which non-linguistic input data is converted into natural language through several intermediate transformations. In contrast, recent neural models for data-to-text generation have been proposed as end-to-end approaches, where the non-linguistic input is rendered in natural language with much less explicit intermediate representations in-between. This study introduces a systematic comparison between neural pipeline and end-to-end data-to-text approaches for the generation of text from RDF triples. Both architectures were implemented making use of state-of-the art deep learning methods as the encoder-decoder Gated-Recurrent Units (GRU) and Transformer. Automatic and human evaluations together with a qualitative analysis suggest that having explicit intermediate steps in the generation process results in better texts than the ones generated by end-to-end approaches. Moreover, the pipeline models generalize better to unseen inputs. Data and code are publicly available.