Whilst translation services are much more effective using trained people, there have been steps to make automated translation software much more intelligent. Recently a new system was created that allows bilingual people to translate written text much faster and with a greater degree of accuracy than before. Experts at American university Stanford have invented a system, which is a combined approach that takes the best from automated software and human intelligence to result in a translation system that could change how we use the service. Computer scientist Christopher Manning, who is a professor of linguistics and computers at Stanford, and graduate Spence Green, have created a dissertation outlining the hybrid approach to translation. Beginnings The origins of this research were borne out from the need to understand foreign languages from post-World War II and Cold War era countries of interest. Computer systems were very successful at breaking codes that were based on specific rules, but human language has many ambiguities that could not successfully be translated by computer systems. Modern world Today, it is a different story. The rapid advancement of technology and the sheer amount of text available on the internet has meant a certain degree of translation can be done online via software such as Google Translate. This provides the general gist, but a bilingual human translator is still needed when a much more precise translation is needed. It’s important to note that human translators still use machines when it comes to translating text. Generally, the person will use machines for a ‘rough’ translation. Their expertise then comes to the fore in the next step as they then edit the text and craft it to create an accurate final version of the text. The proposed Stanford system aims to combine these steps into one seamless operation – fast machine translation and precision human editing combined. The human translator would have an interface that suggests translation for specific terms and keywords that have different meanings. The difference is that system would alter it’s suggestions in real-time as the translator is working through the text. Not a replacement for humans How will this affect the industry? Software that suggests word translation options currently produces almost 22 billion pounds a year in the foreign language market. Human translators can translate almost 3,000 words a day on average. This is around eight single-spaced pages worth of text. Rather than outright replace human interaction, Green says that their system ”…augments the human translator and increases efficiency, accuracy and productivity…” Human translation is still the best and most effective form of translation. Of course, systems like the Stanford approach can only help the industry.
