Take a look at some of the features that have been sold in an automobile, like the automatic transmission, cruise control, auto rain sensing wipers, auto high beam switch, the blind spot auto warning signal, automatic breaking, or automatic parallel parking… All these advancements are helping reach the goal of creating a self-driving automobile, giving a profounder meaning to the term automobile.
Recently Google announced the work that it has been doing in this area; it is impressive to see the progress they made. Google is not designing a self-driving car that can recognize street conditions by machine vision only. The automobile’s computer decision making algorithms rely on a massive amount of data collection about the automobile’s surroundings. Google has painstakingly digitized and indexed 2000 miles of road conditions in California’s Silicon Valley and are planning on crawling and indexing the rest of the streets of the United States, and perhaps the world. Not far-fetched given that they crawled and indexed the worldwide web!
But all these amazing achievements will take a lot of additional work, testing under different weather conditions like rain and snow, legislation changes and time before they reach consumers hands. Even then, it is expected that humans will be given the ability to override the machine when needed.
Similarly, humans have been working on automating translations since the early 50s and much progress has happened since. Although rule-based machine translation showed much promise early on, its reliability proved harder to establish. Then came parsers, segmentors and database tools called translation memory, to help save and reuse existing translations which facilitated the automation of reusing existing translations. About a decade ago, Google started working with Statistical Machine Translation technology to improve the reliability and quality of its machine translation engine. Google stored, indexed and turned all translated data into reusable objects that could produce new translations via statistical methods. Like a self-driving automobile, much language data had to be indexed, stored and analyzed to enable accurate interpretation. But quality is still an issue and human translators continue to be needed to override or replace the machine.
Many are clamoring to create hybrid engines that involve translation memory tools with rule-based machine translation and statistical machine translation engines, coupled with humans to steer and correct the errors to make the process more efficient and accurate. Google just acquired the maker of Word Lens to automate the translation of written text through Google Glass.
Would you trust today a self-driven car to take your kids to school? Would you trust a machine translated manual on how to assemble or operate potentially hazardous apparatus? We live in a very exciting world with daily advancement in technology. In time, both questions may be answered with a resounding yes. Self-driving automobiles could facilitate transportation worldwide. Machine translation engines could improve communication worldwide. Both are excellent motivators to keep the momentum moving forward and together will lead us to an even more exciting new world!
This white paper will guide you through the translation & localization process and shows you how the use of a robust Translation Management System will help alleviate many of the quality translation issues typically encountered in traditional translation processes. Download it for free!Please Share!