Although millions around the world use online machine translation tools to translate many words and texts every day, there is considerable debate about the using such tools by government and official agencies as a reliable means for making their business decisions.
According to Google, the largest online translation service, Google Translate, machine translation is not always accurate, nor is it a substitute for human translators, advising not to rely solely on it for official work. Natural Language Processing specialists say that machine translation still offers ridiculous translations and leads to wrong decisions when dealing with dialects and slang.
This controversy was triggered by the decision of the US Citizenship and Immigration Service to expand the use of online machine translation to examine the status of visa applicants, immigrant applicants, refugees and others by relying on this type of translation to translate the content of their account son social media and various electronic publishing channels.
Text and Context
As the debate intensified, the focus was on “natural language processing,” or NLP, an artificial intelligence technology, as the main component of online machine translation. A number of experts and specialists in the field of academic sector, research and development centers in commercial companies and non-profit organizations, expressed that it is not yet time to rely entirely on machine translation in making decisions regarding the official work of entities and units of States or Governments.
They said that the main reason behind this is that machine translation still deals somewhat superficially with texts, often dealing with the understanding of the meanings of the individual words, the meanings of the sentences and the syntax in their direct form, but they did not go deep enough to understand the context. The full meaning of the text, an issue of cultural and societal dimensions, is most evident when it comes to the translation of local dialects, slang, and literary expressions with rhetorical, rhetorical and semantic dimensions, governed by a diverse “cultural context”. That is why we can see in many cases ridiculous translations that can lead to wrong decisions.
In an e-mail, in response to queries from Business Insider editors who published a report on the issue, Google warned against relying on its popular translation service for complex official tasks, saying it was not intended to be a substitute for human translators or to replace them.
“It would be naive for government officials to rely on online machine translation for their work, such as the government’s reliance on automatic translation to examine social media accounts and social media pages for refugees,” said Douglas Hofstadter, a professor of cognitive science at Indiana University who has studied language and similarity. “It is annoying and probably full of errors, since machine translation services are not designed to analyze nuances or identify slang, so government agencies may misunderstand the true meaning of the content, or miss any actual threat.”
David G. Brizan, a professor of natural language processing at the University of San Francisco, said: “Machine translation services are usually trained using already translated texts, which tend to use more formal speech, such as official UN documents, but this may not be enough, because the language grows, renews and repeats very quickly, especially among young people, which makes it imperative that machine translation develops at the same speed. ” He explained that this is difficult, for several reasons, including the non-textual context of video clips, pictures, groups involved in the conversation and their relationship and cultural references are completely lost when machine translation is used. To address this issue requires what is called «eradicating cultural illiteracy across languages and across generations», which is impossible to achieve at least for now.
ProPublica, a newsroom that produces investigative journalism for the public interest, conducted an investigation into the use of online machine translation in official bodies. During the test, she asked language professors to copy and paste tweets posted in unofficial languages at Google Translate and compare the results with the way they write them, including tweets in several languages. The results were disappointing. For example, a text from Urdu translated with its true meaning, “My father hits a lot but has a lot of love”, as “beating is too big and love too windy”.