Minimum Bayes' Risk Decoding for System Combination of Grammatical Error Correction Systems

TL;DR

This paper introduces a novel Minimum Bayes' Risk decoding method tailored for grammatical error correction systems, improving system combination by aligning decoding with F-score evaluation and demonstrating enhanced performance across datasets.

Contribution

The paper proposes a new MBR loss function directly related to GEC evaluation criteria and an expanded candidate set generation method, advancing system combination techniques.

Findings

MBR decoding improves GEC system combination performance.

Varying reward metrics allows control over precision, recall, and F-score.

Experimental results show state-of-the-art performance on multiple datasets.

Abstract

For sequence-to-sequence tasks it is challenging to combine individual system outputs. Further, there is also often a mismatch between the decoding criterion and the one used for assessment. Minimum Bayes' Risk (MBR) decoding can be used to combine system outputs in a manner that encourages better alignment with the final assessment criterion. This paper examines MBR decoding for Grammatical Error Correction (GEC) systems, where performance is usually evaluated in terms of edits and an associated F-score. Hence, we propose a novel MBR loss function directly linked to this form of criterion. Furthermore, an approach to expand the possible set of candidate sentences is described. This builds on a current max-voting combination scheme, as well as individual edit-level selection. Experiments on three popular GEC datasets and with state-of-the-art GEC systems demonstrate the efficacy of the proposed MBR approach. Additionally, the paper highlights how varying reward metrics within the MBR decoding framework can provide control over precision, recall, and the F-score in combined GEC systems.