eBay is a global shopping powerhouse, with $20.9 billion in quarterly enabled sales (over $80 billion annually), 164 million monthly active users, and nearly 60% of its revenues generated outside the United States.
Since its first article in the eBay tech blog just over six years ago, it has provided insights into how one of the world’s largest commerce platform stays in top performance: in-depth discussions of data and application security, methodologies for continuous integration (CI), machine learning, big data, and Hadoop, user interface and user experience, and in-the-trenches views of developer tools and open source resources.
This week, Juan Rowda of eBay’s Machine Translation team began a tour of how eBay uses a myriad of tools to analyze and visualize localization quality. In Part I of his blog, Visualizing Machine Translation Quality Data, he focused on how to effectively put all the data eBay has on its localization efforts into measurable, visually-understandable, and actionable insights. Tools like Excel, TAUS DQF, and Tableau allow the visualization of data critical for localization decision-making.
Perhaps the issue is that too much information can be hard to make sense out of and may even feel overwhelming. That is, precisely, the advantage of data visualization. — Juan Rowda, eBay
The particular domain of focus for this part was vendor analysis, particularly for Machine Translation Post-Editing (MTPE). [Clarification: e2f was not one of the vendors used by eBay for this analysis.]
In Rowda’s article, quality data was broken out by content type (titles or descriptions), by vendor (and even individual reviewer), by language, and by defect type. Were these issues in mistranslation? Untranslated words? Additions or omissions? Terminology? Or was the MT content itself not available or inconsistent?
The article shows a depth of quality management consistent with a world-class operation. We’re eagerly awaiting part two of his blog!
What are your thoughts? How do you manage and visualize quality at your organization? How does eBay’s visualization methods compare to and inform your own? We’d love to hear! Email us at [email protected], and let us know your thoughts.