Check the preview of 2nd version of this platform
being developed by the open MLCommons taskforce on automation and reproducibility
as a free, open-source and technology-agnostic on-prem platform.
Crowdsourcing for Evaluating Machine Translation Quality
lib:de2d022b447cd9df (v1.0.0)
Authors: Shinsuke Goto,Donghui Lin,Toru Ishida
Where published: LREC 2014 5
Document: PDF DOI
Abstract URL: https://www.aclweb.org/anthology/L14-1592/
The recent popularity of machine translation has increased the demand for the evaluation of translations. However, the traditional evaluation approach, manual checking by a bilingual professional, is too expensive and too slow. In this study, we confirm the feasibility of crowdsourcing by analyzing the accuracy of crowdsourcing translation evaluations. We compare crowdsourcing scores to professional scores with regard to three metrics: translation-score, sentence-score, and system-score. A Chinese to English translation evaluation task was designed using around the NTCIR-9 PATENT parallel corpus with the goal being 5-range evaluations of adequacy and fluency. The experiment shows that the average score of crowdsource workers well matches professional evaluation results. The system-score comparison strongly indicates that crowdsourcing can be used to find the best translation system given the input of 10 source sentence.
Where published: LREC 2014 5
Document: PDF DOI
Abstract URL: https://www.aclweb.org/anthology/L14-1592/
The recent popularity of machine translation has increased the demand for the evaluation of translations. However, the traditional evaluation approach, manual checking by a bilingual professional, is too expensive and too slow. In this study, we confirm the feasibility of crowdsourcing by analyzing the accuracy of crowdsourcing translation evaluations. We compare crowdsourcing scores to professional scores with regard to three metrics: translation-score, sentence-score, and system-score. A Chinese to English translation evaluation task was designed using around the NTCIR-9 PATENT parallel corpus with the goal being 5-range evaluations of adequacy and fluency. The experiment shows that the average score of crowdsource workers well matches professional evaluation results. The system-score comparison strongly indicates that crowdsourcing can be used to find the best translation system given the input of 10 source sentence.
Relevant initiatives
Related knowledge about this paper
Reproduced results (crowd-benchmarking and competitions)
Artifact and reproducibility checklists
- ACM/cTuning artifact appendix and reproducibility checklist
- NeurIPS reproducibility checklist
- SIGPLAN's checklist for empirical evaluation
- Collective Knowledge (organizing research projects based on FAIR principles)