Sentence-level quality estimation by predicting HTER as a multi-component metric

TL;DR

This paper proposes a machine learning approach that predicts individual post-editing operations to estimate sentence-level HTER scores, improving accuracy without extensive feature engineering.

Contribution

It introduces a multi-output neural model that jointly predicts editing operations, enabling better HTER estimation and correction of invalid predictions.

Findings

Multi-layer perceptron with 4 outputs improves HTER prediction accuracy.

Joint prediction of editing operations enhances estimation robustness.

Model allows correction of invalid HTER predictions.

Abstract

This submission investigates alternative machine learning models for predicting the HTER score on the sentence level. Instead of directly predicting the HTER score, we suggest a model that jointly predicts the amount of the 4 distinct post-editing operations, which are then used to calculate the HTER score. This also gives the possibility to correct invalid (e.g. negative) predicted values prior to the calculation of the HTER score. Without any feature exploration, a multi-layer perceptron with 4 outputs yields small but significant improvements over the baseline.