One of the biggest pitfalls of MT is its inability to translate the non-standard language with the same precision as the default language. Heuristics or statistics MT takes entries from different sources in the standard form of a language. By nature, rule-based translation does not include generic non-standard uses. The result is errors in translation from a popular language or a familiar language. Restrictions on translation from casual language pose problems when using machine translations on mobile devices. The designated entities must first be identified in the text. Otherwise, they may be wrongly translated as common nouns, which would probably not affect the UEBL`s assessment of translation, but would alter the human legibility of the text.  They may be omitted in the translation, which would also have consequences on the readability and message of the text. Machine translation can use a method based on dictionary entries, which means that words are translated as they are, by a dictionary.
Franz Josef Och (the future director of translation development at Google) won the DARPA SPEED MT competition (2003).  Other innovations at that time were MOSES, the engine of open source statistics MT (2007), a text/SMS translation service for mobile phones in Japan (2008) and a mobile phone with integrated voice translation to language functions for English, Japanese and Chinese (2009). In 2012, Google announced that Google Translate translates about enough text to fill 1 million books in one day. Transliteration involves the search for letters in the target language that best correspond to the name in the original language. However, this has been mentioned as a sometimes deterioration in the quality of translation.  For « Southern California, » the first word should be translated directly, while the second one should be transliterated. Machines often transverse both because they have treated them as a single entity. Words like this are difficult for machine translators to process, even those with a transliterator component. SYSTRAN, which in the 1960s « under U.S. government contracts was the site of the pioneers of the field, was used by Xerox for the translation of technical manuals (1978). From the late 1980s, when computing power increased and became cheaper, statistical models for machine translation became more interested. MT became more and more popular after the advent of computers.
 The first systRAN implementation system was set up in 1988 by the online service of the French post office Minitel.  Several MT companies were also established, including Trados (1984), which was the first to develop and commercialize translation memory technology (1989). The first MT trading system for Russian/English/German-Ukrainian was developed at Kharkov State University (1991). Machine translation, sometimes referred to as MT (not to be confused with computer-assisted translation, machine human translation or interactive translation) is a sub-domain of computer linguistics that studies the use of software to translate text or language from one language to another. Only the originals are copyrighted, so some scientists claim that the results of machine translation are not entitled to copyright protection, because MT does not imply any creativity.  The right to challenge is for a derivative work; the author of the original work in the original language does not lose his rights when a book is translated: a translator must have permission to publish a translation. Part of the use of a multiparallel body, i.e. a body of text translated into 3 or more languages, has been carried out. With these methods, a text translated into two or more languages can be used in combination to allow a more accurate translation into a third language, compared to only one of these source languages.    Better output quality can also be achieved through human intervention: some systems can translate more accurately when the user has clearly identified