Crowdsourcing is recently used to automate complex tasks when computational systems alone fail. In this project, we investigate how humans can effectively contribute to automate natural language translation. The envisioned goal is a hybrid machine translation service that incrementally adapts machine translation models to new domains by employing human computation to make machine translation more competitive. Therefore, we investigate efficient ways for domain adoption of neural machine translation systems using crowd-generated input data.

Poster at Collective Intelligence 2018

In this work, we investigated (1) whether a (paid) crowd, that is acquired from a multilingual website’s community, is capable of translating coherent content from English to their mother tongue; and (2) in which cases state-of-the-art machine translation models can compete with human translations for automation in order to reduce task completion times and costs.

References

  • Michael Barz, Tim Polzehl, Daniel Sonntag: Towards Hybrid Human-Machine Translation Services. EasyChair Preprint no. 333, 2018.
  • Michael Barz, Neslihan Büyükdemircioglu, Rikhu Prasad Surya, Tim Polzehl, Daniel Sonntag: Device-Type Influence in Crowd-based Natural Language Translation Tasks. In: Checco, Alessandro; Demartini, Gianluca; Gadiraju, Ujwal; Sarasua, Cristina; Aroyo, Lora; Dumitrache, Anca; Paritosh, Praveen; Quinn, Alex; Welty, Chris (Ed.): 1st Workshop on Subjectivity, Ambiguity and Disagreement (SAD) in Crowdsourcing 2018, and CrowdBias'18: Disentangling the Relation Between Crowdsourcing and Bias Management, pp. 93–97, Aachen, 2018, ISSN: 1613-0073.

Contact

Michael Barz and Marimuthu Kalimuthu