NMTScore: A Multilingual Analysis of Translation-based Text Similarity Measures

TL;DR

This paper introduces NMTScore, a multilingual framework for analyzing translation-based text similarity measures, demonstrating their effectiveness and robustness compared to traditional methods across various languages and tasks.

Contribution

The paper presents a comprehensive analysis of translation-based similarity measures within a multilingual NMT framework and releases the NMTScore library for broader use.

Findings

Translation-based measures are competitive with sentence embeddings in paraphrase detection.

They are more robust against adversarial and multilingual input when normalized properly.

Translation-based measures correlate well with human judgments in data-to-text evaluation across multiple languages.

Abstract

Being able to rank the similarity of short text segments is an interesting bonus feature of neural machine translation. Translation-based similarity measures include direct and pivot translation probability, as well as translation cross-likelihood, which has not been studied so far. We analyze these measures in the common framework of multilingual NMT, releasing the NMTScore library (available at https://github.com/ZurichNLP/nmtscore). Compared to baselines such as sentence embeddings, translation-based measures prove competitive in paraphrase identification and are more robust against adversarial or multilingual input, especially if proper normalization is applied. When used for reference-based evaluation of data-to-text generation in 2 tasks and 17 languages, translation-based measures show a relatively high correlation to human judgments.