Enhancing machine translation with crowdsourced keyword highlighting
Mei-Hua Pan, Hao‐Chuan Wang · 2014
Machine translation (MT) has the potential to bridge the language barrier in multilingual community, but its utility to support cross-lingual communication is often limited by its quality. Recent studies have shown the supporting effect of keyword highlighting on translation comprehension. What's missing is a way to identify and highlight keywords in translations efficiently and reliably. In this paper, we investigate crowdsourcing strategies for keyword highlighting. We compare three methods that involve human workers to highlight keywords in English-to-Chinese translations: English-only for adding keyword highlights on the original English sentences, Chinese-only for adding keyword highlights on the translated sentences, and Bilingual that allows people to choose keywords in either the original or the translated sentences. Results show that adding highlights to translated sentences (Chinese-only) is the fastest but doesn't improve the comprehensibility of translation. Highlighting both sentences (Bilingual) helps improve the understandability of hard-to-understand translations.