Google Translate Disagreement
The results are not absolute for two reasons. First, the choice of expressions was arbitrary. Another selection of English phrases would generate different numerical results in each language. for example, notes would probably fall if more idioms like the scrambled “like a bat from hell” were received. However, the relative results would remain roughly the same; A language with high scores for my test game should perform very well with other phrases, while a language with low scores would have equally low results with other entries. Second, most language results show the subjective opinion of a single critic. It could be quite argued that more criticisms per language would produce more reliable data. Several languages had several criticisms, and disagreements of inter-annotators were generally minor, with a handful of entries per language judged to be well versed marginal, or marginal versus false. A single entry, considered good and marginal by different evaluators, does not change Tarzan`s score and changes the bard score by 2.5 points, and a disagreement between marginal and erroneous changes to both Tarzan and Bard by 2.5.
I cut notes where the thieves disagreed, and kept the disagreements visible in the publication of the public data. Based on the annoyance values of the intermediate annotator I discovered, it is recommended that the reader place mental error bars of ± 10 to get the recorded score. I found that evaluators who spoke English at a very high level tended to score higher than less qualified people. For example, Bard`s assessment for Indonesian by a spokesperson who worked in a European Google office (not for GT) was 10 points higher than that of an evaluator who has always lived in Indonesia. In some cases, respondents tried to re-evaluate the survey by replacing the paraphrases provided by the original sentence and assessing whether the translation translated the real words – for example, translated “steam” rather than whether it reflected the idea of exhaustion. The most successful language in this study is Afrikaans, which has very close ancestors with English and has only drifted from Dutch for the past 300 years, while interacting intensely with English in South Africa. African evaluators understood the figurative meanings of several sentences translated word for word, and these literal translations did not resonate with locs of other languages. On the other hand, some English metaphors have been adopted in other languages. For example, “the closet” was understood as a hidden sexual identity in some places where the local gay community imported this meaning to its own word for closet, or, in Latin American Spanish, actually borrowed the word English. This study raises more questions than it answers – z.B. Xhosa and Zulu intersect with English and Afrikaans in the same South African landscape, but close to the ground, and this study is not able to determine how much this difference is due to the difference in exposure of the evaluators in English, to the similarity between Afrikaans and English, which does not exist for the two indigenous African languages. translation limitations proposed by GT for Xhosa and Zulu.