Post-editing effort in machine translation is influenced by source text complexity and editor behaviour.

Category: Human Factors · Effect: Moderate effect · Year: 2010

The amount of effort required to post-edit machine-translated text is not solely dependent on the machine's output but is significantly affected by the characteristics of the original text and the specific editing strategies employed by the human editor.

Design Takeaway

When designing or implementing systems that involve human-machine collaboration, consider the cognitive load and decision-making processes of the human operator, as these are as crucial as the machine's performance.

Why It Matters

Understanding these influencing factors is critical for optimizing workflows that integrate machine translation with human oversight. This knowledge can lead to more efficient resource allocation, improved quality control, and better training for post-editors, ultimately enhancing the productivity and output of translation services.

Key Finding

The complexity of the original text and how editors approach the task both play a significant role in how much work is needed to correct machine translations.

Key Findings

Research Evidence

Aim: To investigate how source text characteristics and post-editing behaviour influence the effort required for post-editing machine-translated text in a commercial setting.

Method: Mixed-methods research (quantitative and qualitative observation and analysis)

Procedure: Professional Japanese post-editors' work was observed in real-life industrial settings. Data on the amount of editing, source text features, and editor actions were collected and analyzed both statistically and qualitatively.

Context: Commercial translation services utilizing machine translation and human post-editing.

Design Principle

Optimize human-machine collaboration by accounting for human cognitive factors and task-specific contextual influences.

How to Apply

When developing or refining translation workflows, analyze the source material for inherent complexities and observe post-editors to identify common challenges and effective strategies. Use this insight to tailor tools and training.

Limitations

The study focused on Japanese post-editors, so findings may not be universally generalizable to all language pairs or cultural contexts. The specific machine translation system used was not detailed.

Student Guide (IB Design Technology)

Simple Explanation: Even when a computer translates something, how hard a person has to work to fix it depends on how tricky the original text was and how the person chooses to edit it.

Why This Matters: This research highlights that user performance is not just about the tool but also about the user's interaction with the tool and the task's inherent difficulties, which is a key consideration in many design projects.

Critical Thinking: How might the 'intertwined manner' of factors affecting post-editing effort be further deconstructed to identify specific design interventions for different types of source text complexity?

IA-Ready Paragraph: This research by Tatsumi (2010) demonstrates that in collaborative human-machine tasks, such as post-editing machine translations, the effort required from the human operator is significantly influenced by both the characteristics of the input material (e.g., sentence structure, specialized terminology) and the operator's specific editing behaviours and strategies. This suggests that design interventions should not only focus on optimizing the machine's output but also on supporting the human user in navigating task complexity and employing efficient workflows.

Project Tips

How to Use in IA

Examiner Tips

Independent Variable: ["Source text characteristics (sentence structure, document component types, product-specific terms)","Post-editing behaviour/patterns"]

Dependent Variable: Amount of post-editing effort

Controlled Variables: ["Commercial setting","Professional post-editors","Japanese language"]

Strengths

Critical Questions

Extended Essay Application

Source

Post-editing machine translated text in a commercial setting: Observation and statistical analysis · Dublin City University Open Access Institutional Repository (Dublin City University) · 2010