PragExTra: A Multilingual Corpus of Pragmatic Explicitation in Translation

TL;DR

PragExTra introduces the first multilingual corpus and detection framework for pragmatic explicitation in translation, combining automatic candidate extraction via null alignments with a human-in-the-loop annotation pipeline. A multilingual BERT classifier, trained through an extensive active-learning cycle, achieves high cross-lingual performance (up to $0.88$ accuracy and $0.82$ F1) and demonstrates that entity- and system-level explicitation are common phenomena across domains. The work reveals systematic cross-linguistic patterns in explicitation, highlights the value of active learning for rare, culturally motivated additions, and provides a resource for building culturally aware machine translation. While promising, it acknowledges limitations in language coverage and context, suggesting future extensions to multimodal signals and broader domains.

Abstract

Translators often enrich texts with background details that make implicit cultural meanings explicit for new audiences. This phenomenon, known as pragmatic explicitation, has been widely discussed in translation theory but rarely modeled computationally. We introduce PragExTra, the first multilingual corpus and detection framework for pragmatic explicitation. The corpus covers eight language pairs from TED-Multi and Europarl and includes additions such as entity descriptions, measurement conversions, and translator remarks. We identify candidate explicitation cases through null alignments and refined using active learning with human annotation. Our results show that entity and system-level explicitations are most frequent, and that active learning improves classifier accuracy by 7-8 percentage points, achieving up to 0.88 accuracy and 0.82 F1 across languages. PragExTra establishes pragmatic explicitation as a measurable, cross-linguistic phenomenon and takes a step towards building culturally aware machine translation. Keywords: translation, multilingualism, explicitation

Figures (2)

• Figure 1: Main Findings. Pragmatic expliciations introduce cultural or background knowledge that is not present in the source text to make the text more accessible to the target audience. To model this, we create seed data for a multilingual-BERT using null-alignments and active learning based on monolingual classifiers (top). The trained model predicts explicitation labels across ten language pairs (bottom), resulting in the PragExTra corpus of 2,700 sentence pairs exhibiting pragmatic explicitation.
• Figure 2: Classifier Performance. Results averaged over 5 random seeds.