In 2013, eBay announced a big focus on growth in emerging markets for its marketplace, in particular, Russia, Brazil and China. Some of this growth can be enabled through localization, but the marketplace has tested a more technical approach — with machine translation — with its first big expansion effort in Russia.
To spearhead these efforts, eBay brought on machine translation expert Hassan Sawaf, a data scientist whose career spans more than 20 years in speech recognition and human translation technologies. He’s also the patent holder on hybrid machine translation, a system and method for using machine translation to translate from one language into another.
As Sawaf explains to us, language translation can be a source of friction between buyers and sellers on eBay, and his goal was to go beyond word-for-word translation into what he calls context translation. This means that Sawaf is helping build engines that ‘learn’ from context of the data (like item descriptions) rather than just more standard word-by-word translations.
Here’s the current problem eBay faces in emerging countries like Russia. eBay is trying to curate inventory from a global base of sellers and surface this to buyers in emerging eBay markets based on what ships to them in their respective countries. A Russian user can go to the localized version of eBay and see all products that are listed in Russian. When they are inputting search terms in Russian, this engine will produce search results of listings that match the query in Russian. But the Russian user’s query will not be able to see posts that match their query that were written by sellers listing in English. In order to access English listings, which do represent a considerable number of the listings on eBay’s platform, Sawaf explains, the Russian user would have to input the query on eBay in English.
“Machine translation normalizes this,” says Sawaf.
For the past year, Sawaf and his team of 14-15 data scientists and engineers have built a technology that allows Russian users to search in Russian, but be able to return queries with English listings that match. Sawaf’s technology takes it one step further, and will determine that a ‘purse’ in an item description, also refers to ‘bag,’ or ‘item,’ or ‘piece,’ providing a more accurate representation of the item in the Russian language. Sawaf says that the search technology, which just launched a month ago, returns signficantly more results for Russian eBay users. And twice the amount of users in Russia are inputting search queries in their native language. It’s unclear how this has translated into an increase sales and transactions in the marketplace, which is the ultimate goal.
Now that Sawaf has this scalable infrastructure in place, the team will be expanding this to other languages in emerging markets. We hear that the team is tackling Spanish and Portuguese, focusing on Brazil and Latin America.
Some e-commerce companies outsource some of the machine translation work to third-party providers, and eBay considered this, but Sawaf tells us that, “to develop the best, you have to do it on your own.” Plus, there is some third-party user data that eBay did not want to share with third parties.
We’re told that along with this improvement in search, eBay is also attempting to make improvements for Russian users with payments, shipping (an area that has faced some challenges in Russia) and other services in these markets. Another interesting takeaway is that eBay’s ultimate goal in these emerging markets is to offer more in B2C selling on the platforms, and find ways to get more local businesses selling on eBay online.
While eBay says Russia has been a No. 1 priority, it will be interesting to see how the improvement in technologies translates into actual sales. And how eBay’s localization performance in other markets will also be a way in which we can grade whether machine learning is working.
It’s worth noting that India hasn’t been a success story for eBay. As TechCrunch writer Pankaj Mishra wrote recently, despite entering India early, eBay has not become the dominant leader. eBay backed local e-commerce marketplace SnapDeal in a possible effort to make up for past mistakes. SnapDeal is six to seven times bigger than eBay in volume of business. Interestingly, Sawaf and eBay didn’t really mention India in the localization and machine-learning efforts.
But if eBay’s machine-learning technology can translate into an increase of sales in emerging e-commerce markets like Latin America and Russia, this could represent billions in new revenue. Stay tuned.