News - Fresh Data Archive Article

Return to Fresh Data Blog
Return to Fresh Data Archive

China Post to Upgrade Postal Address Translation System

Date: June, 2010 --

If you have a Chinese file of customers or prospects, you probably have it in the Roman alphabet, which is perfectly natural, and it’s reasonably functional. However, as anyone who has been in China knows, the use of the Roman alphabet in daily life, including within the postal service, is very rare. After all, would we expect the USPS to deliver letters written in a Chinese translation of our addresses?

In fact, if you visit any of the 39 International Exchange Offices (IEC) of China Post, you will see an amazing sight. But you better go soon, because the sight is going to disappear.

In each IEC there is a large room which receives in-bound mail and which contains desks arranged in an efficient logistic order that accommodates some 10 or 15 specially trained experts. You will find them hard at work at desks piled high with incoming mail of all sorts: cards, letters, newspapers, magazines, parcels and the entire gamut of postal traffic.

These folks read the Roman alphabet address as written on the inbound mail and rewrite it into Chinese characters. Then it is sent for sortation. This is very costly and the training time for a proficient address translator is over one year. Moreover, the throughput with a full staff of three shifts is only about 30,000 pieces per day. Even so, this is an amazing amount of work. And remember, none of that work involves address validation or COA.

This is about to change for the benefit of China Post, its customers, and all foreign mailers.

At an important meeting in Switzerland of the Universal Postal Union's Address Working Group in April, China Post presented its new address translation system, which is now being installed in those 39 IEC's and some 31 other provincial offices.

The system is the result of a great deal of complex work in optical character recognition, translation, and automatic learning. Inbound letters are fed into one of the two models of a deceptively simple-looking machine, either an automatic or semi-automatic version.

An optical character recognition system captures the Roman alphabet address and extracts the English keywords. These keywords are then translated by reference to a Chinese keyword database. Some checking of the address is carried out by reference to China Post’s address database and the subsequent translation is then printed onto the envelope.

While this sounds like a simple and straightforward operation, there are many different ways to "transliterate" Chinese characters into English spelling, and of course, as in any country, there are new addresses being created in China all the time. Because of this, the system contains an "intelligent learning platform" which increases the knowledgebase of the system.

In short, the system is self-educating. It contains an algorithm based on frequency which can generate new Chinese keywords, or properly identify the existing Chinese character when it confronts a new English spelling. It will also identify new addresses and add these to its knowledgebase. Moreover, this system currently boasts an accuracy rate of 90%.

Where speed is concerned, the system is fast, automatically processing 3,000 letters an hour per automatic machine, while somewhat slower with a non-automatic machine that handles flats and parcels.

Naturally, your promotions will go through this process, and for it to be as accurate as possible, be sure to have Data Services, Inc. clean up the file so the English spelling of your addresses will be as standard as is possible to assure accurate "machine translation", and thus delivery.