Enhancing SPARQL Generation by Triplet-order-sensitive Pre-training

TL;DR

This work proposes an additional pre-training stage with a new objective, Triplet Order Correction (TOC), along with the commonly used Masked Language Modeling (MLM), to collectively enhance the model's sensitivity to triplet order and SPARQL syntax.

Abstract

Semantic parsing that translates natural language queries to SPARQL is of great importance for Knowledge Graph Question Answering (KGQA) systems. Although pre-trained language models like T5 have achieved significant success in the Text-to-SPARQL task, their generated outputs still exhibit notable errors specific to the SPARQL language, such as triplet flips. To address this challenge and further improve the performance, we propose an additional pre-training stage with a new objective, Triplet Order Correction (TOC), along with the commonly used Masked Language Modeling (MLM), to collectively enhance the model's sensitivity to triplet order and SPARQL syntax. Our method achieves state-of-the-art performances on three widely-used benchmarks.

Figures (2)

• Figure 1: The overview of our approach. The TosT5 model first undergoes the triplet-order-sensitive pre-training stage and then is fine-tuned on the downstream task.
• Figure 2: Error analysis on models' prediction. "tfe" is short for "triplet-flip error", and "te" is short for "triplet error".