Translation is a complicated job. You would find multiple companies, including some big names, claiming how they can translate a website into another language in a jiffy. Jiffy part might be accurate but what comes out after the translation may look like some words put together by a kid who is just learning to talk. For instance, one funny example to quote after using a certain translation tool, “I don’t care” got translated into “What goes of your father?” A link to “Twitter” got transformed into “singing bird”. As you can imagine, results can get disastrous if translation is too literal and done without an appropriate context….in short too mechanical and without the human touch.
Machine Translation constitutes of rule based algorithms and statistical methods. Although these engines are on their way to becoming more sophisticated than before, essentially, they are based on dictionaries and algorithms. Usually words are substituted with the equivalent in the other language with attention to sentence formation. For example, a sentence in Hindi does not quite follow the same grammatical rules and positioning as in English. Hence verbatim translation is not very effective and that is where Machine Translation falls short of users’ expectations.
In certain applications, however, e.g., product descriptions written in a standard, controlled language, or a legal document, a dictionary-based machine-translation system works quite well. Save for proofreading and quality inspection, minimal human intervention is required. For these unambiguous translation projects, a machine tool can be substituted sporadically but quality is always a concern. Also, domain knowledge is very important as well. Depending on the context, a synonymous dictionary translation will not produce the intended effect. For example “Stab” synonym “penetrate” can change the meaning of a sentence completely. Another risk associated with machine translation is incorrect spellings which can be somewhat mitigated depending on the domain knowledge of the translation experts.