How eBay Leverages Data Visualization in Machine Translation

Online retailers such as eBay are making the case for using data visualization to sort through big data to improve automated language translation in e-commerce markets where platform operators are striving to bring together buyers and sellers around the world.

Data scientists at eBay make the case that data visualization has emerged as one of the best ways to “assimulate information in a very natural way.” The retailer is focusing on three different tools for visualizing “translation quality” data: Tableau, TAUS Dynamic Quality Framework and Excel.

These and other data visualization tools help reveal patterns and trends, making them “easier to spot and, sometimes, even obvious,” Juan Rowda, a machine learning specialist at eBay noted in a technology blog. “Correlations may pop up and give you much-needed business or strategic advantages, allowing you to effectively act on your information.”

Among the challenges for global retailers such as eBay is “localization” of content so buyers and sellers can communicate down to the level of alphanumeric product specifications and company shorthand like NIB (“new in box”). “It looks like the localization industry is really not doing a lot to harness all this information,” Rowda noted. “The industry seems to be missing out on this and is not fully leveraging, or at least trying to understand, all these wonderful bits and pieces of information that are generated every day.

Rowda said eBay uses data visualization to track the quality of its machine translation output for different language combinations as well as details on issues found with its machine translations. It also tracks the performance of vendors using its platform and any issues they encountered related to translations.

The company’s overview of data visualization tools starts with Tableau, which it uses to generate visualizations on, for example, how vendors are performing based on content types used for search queries, including product titles and descriptions.

Tools like Tableau revealed that language “mistranslations” are a common problem, and that they occurred most frequently when translating from the French. In another example, “post-editors” who double check machine translations might do well with product descriptions but fall short on product titles.

In using data visualization tools such as Tableau, the eBay data scientist stressed the importance of being selective rather than cramming columns and lines into a data visualization. “Carefully plan the data points you want to include, assessing if they contribute or not to the main purpose of your message,” Rowda stressed. “This can be difficult, especially if you have too much information – you may feel tempted to add information that might not add any value at all.”

Separately, the online retailer said machine translation is being expanded to cover product reviews. Those algorithms use statistical translation models tailored for the e-commerce market. “For instance, neural network architectures for machine translation allow us to make use of wider context and eBay’s structured metadata in the most effective way,” noted Evgeny Matusov, eBay’s senior manager of machine translation science.

Recent items:

Search Engines Get a Machine Learning Boost

Hadoop Gets a Security Boost From eBay’s Eagle