The need for localization and especially translation continue to grow as the global nature of business sets in. Online selling platform eBay has called in the help of machine translation experts to help boost sales in emerging markets.
According to Leena Rao, eBay has been looking into the concept of machine translation to improve the customer experience for their Russian users. In her article on TechCrunch, she tells us all about these new developments.
Last year, eBay stated that it would target growth in emerging markets. The company mainly focussed on Russia, Brazil and China and partly used localization to expand their market shares in these countries.
However, eBay also called in machine translation for expansion. In order to become a bigger player on the Russian market, Rao says, eBay hired the very experienced machine translation expert Hassan Sawaf. Sawaf is a data scientist who is mainly concerned with speech recognition and human translation technologies. Moreover, according to Rao, he also is the owner of a patent on hybrid machine translation.
Sawaf believes the language barrier between buyers and sellers on eBay can be troublesome. In order to solve this issue, Sawaf aimed for “context translation.” In other words, he is aimed to create a system that can ‘learn’ from the context in which the data is in, item descriptions for example, instead of making word-by-word-translations.
According to Rao, eBay is experiencing the following problem on emerging markets such as Russia: if Russian customers enter a Russian search term in eBay’s search engine, the results will only display items that include that Russian word. This might leave out international sellers that ship to Russia, but only list their items in English. Machine translation is a great way to quash this problem, Sawaf says: it will ensure Russian visitors see all products that are shipped to their country.
Sawaf and his team of about 15 data scientists and engineers have been working on their machine translation system for a year. This system produces English listings for Russian queries, but that is not the end of it, Rao says. It will also be able to note that the word “purse” has a relation with words such as “bag” or “item,” resulting in a accurate listings. The technology has been launched about a month ago and according to Sawaf, Russian users now get more elaborate search results than ever before. However, Rao says, even though the ultimate goal of the software is to increase the number of transactions and sales, it is still unknown whether this has actually happened.
Sawaf’s system will also be implemented in other localized versions of the eBay website such as Brazil and Latin America, Rao says. Even though eBay considered to outsource the machine translation work, Sawaf thinks it is always better to develop tools yourself as this will give the best results. In addition, eBay opted for his help because the company did not want to share all of its data with third parties.
According to Rao, eBay is trying to improve their website for their Russian consumers in other ways as well, for example by implementing new payment and shipping methods. She also finds it interesting that eBay’s main aim in these markets is to offer more B2C selling and to attract more local businesses. Rao is keen to find out whether the technological improvements eBay has made will affect actual sales. Finally, she also states that eBay’s experiment is a great way to see whether machine learning is an effective translation method.
Rao does like to point out that eBay’s previous adventure in India has not been very successful. In an earlier article on TechCrunch, Pankaj Mishra revealed that even though eBay was one of the first players on the market, it has not become the market’s leader.
Sawaf, did not mention India at all in his talks about localization and machine-learning translation tools. However, Rao believes eBay could make millions if their new translation technology is able to boost sales in emerging markets such as Russia and Latin America.