Accuracy, Readability and Acceptability in The Translation of Android Xiaomi Redmi Note 4

____________________ Abstract ___________________________________________________________________ Android translation(s) were translated by multiple people/agencies during multiple period of times, thus they are very prone to accuracy, readability and acceptability errors. Based on that reason, I conducted this research to (1) analyze and find out the accuracy of the translation in Xiaomi Redmi Note 4 smartphone, (2) readability of the translation of the smartphone, and (3) the acceptability of the translation. The result of this study showed that all of the accuracy, readability and acceptability of the original translation showed worse results than the modified translation. For accuracy, Goff-Kfouri‘s and Nababan‘s rubric both showed better result for the modified translation. For readability by the translators, the average of original translation got 2.3 (negative) and modified got 3.8 (positive); whereas the end-users rated 2.4 (negative) and 3.8 (positive) respectively. For acceptability, the average of original translation by the translators was rated 2.6 (negative) and the modified translation got 3.7 (barely positive);


INTRODUCTION
Translators must not only understand the two languages; i.e. source and target languages, but also the other competences such as communication competence, discourse, and context of situation. Translators must be fully capable of conveying correct messages from SL to TL, so the people in TL environment can understand the message fully (communicative competence). They must be able to convey messages that might seem invisible, only exist in either TL or SL's culture, and other pragmatic messages (discourse). They must also see the whole context of situation of SL, so it will not be misunderstood by the people in TL environment. And last, translators must be able to utilize correct linguistic elements in TL environment, such as language style, practicality, readability and literacy.
Commercial mobile phones, from earlier apperance in Indonesia around year 2002, have been translated to Indonesian language. Although it is unclear which phone was translated the first, Indonesian translation was found in Nokia, Motorola, Siemens and Sony Ericsson. The three brands were among the first to be commercially available in Indonesian language in a relatively affordable prices. Albeit slight differences, there are similarities on the translations among the brands. Among others, -messages‖ is translated as -pesan‖, -call‖ is -panggilan‖, and -setting‖ is -Setelan‖. Android translations, however, are different in each brand, even in each smartphone product. Each brand employs individual or even several translation agencies to translate their products. Even then, one agency may not be employed continuously by a brand. This causes potential inconsistencies among phones in a brand, even in one same phone. Not only that the translations could be inconsistent, but the quality could be compromised as well due to different translation agency having different quality control standards.
This research is focused on the accuracy, readability and acceptability of the translation of Xiaomi Redmi Note 4 smartphone. Accuracy, taken from Molina & Albir (2002), is defined as whether a translation possesses a certain degree of correctness according to certain requirements. For example, Nababan (2012) sets out three degrees of accurateness: Accurate, Less Accurate, and Not Accurate; each with corresponding descriptions on the assessment criteria. Skopos on the other hand does not regard accurateness as important. Due to the lack of equivalence, and the focus on whether the reader understand or not makes accuracy less meaningful than correctly placed sentence structure. This, in essence, makes the assessment of translation solely based on the general understanding of target readers. Larson (1984), however, added that accuracy means that the information between the source and target is simply correct. Any changes, addition or deletions is conducted after comparing source and target text, also thinking how good the translation will be in the target situation. In short, Larson's theory on accuracy is aligned well with Skopos, although dynamic equivalence still plays a huge part in the translation process.

Translation
Catford (1965) referred translation as a process of substituting a language into another language. Larson (1984) explained further that a translation is not only a change of language, but also a transfer of the meaning, by the means of a change in semantics, a constant transfer between source and target language, and a clear reexpression of a source text into the target language. Larson claimed that only the while the form changes, while other messages should be encoded and re-expressed in target language in a proper equivalence manner. Newmark (1989) simplifies the term back by arguing that translation is merely a process of converting any utterance of any source language to the target language. Larson (1984) further divides translation into two, form-based and meaningbased. Form-based translation, as the name suggests, attempts to translate a text solely based on the form of the source text. Meaning-based translation is the contrary, where a translator attempts to follow the source language.
Regarding translation strategy, Cohen (1984), Loescher (1991) and Krings (1986) argued that translation strategy is a translator's conscious attempts and plans in solving translation problems, as well as translator's ability to distinguish the correct or incorrect translating methods. Bell (1998), similar to the above-mentioned theories, adds a differentiation for the strategies: global and local. Global strategy means that it deals with translation strategy as a whole process-from start to finish, and local means that it deals with certain problems during certain smaller parts in the translation. In relation to this research, skopos deals with smaller translations, hence it used the local strategy.

Accuracy, Readability, and Acceptability
Accuracy (Molina & Albir, 2002), is defined as whether a translation possesses a certain degree of correctness according to certain requirements. Nababan (2012) sets out three degrees of accurateness: Accurate, Less Accurate, and Not Accurate; each with corresponding descriptions on the assessment criteria. Nida (1964) stated that translation is a set of procedures in order to create a meaning in a language target reader can understand. Larson (1984), added that accuracy means that the information between the source and target is simply correct. Any changes, addition or deletions is conducted after comparing source and target text, also thinking how good the translation will be in the target situation. In short, Larson's theory on accuracy is aligned well with Skopos theory.
Readability is a way of finding the best translation methods and techniques to fit with the source text so the readers can easily understand them (Dubay, 2004). Hartono (2017) further expands Larson‗s (1984) definition by adding that readability can be determined by diction, structure and organization of sentences, spelling and even punctuations. Yolanda and Yuliasri (2016) argues that readability prioritizes the translation to sound natural instead of being simply correct. Nababan (2012) also added that in written translation, readability shows how much a text is easily understood by the target readers.
Acceptability is how a translated text correctly reflects the target culture, norm and linguistic rules (Nababan, 2012). In essence, the translation must comply with the locally acceptable rules; in terms of language style, whether slang is involved or not, even multimeaning words in certain cultures. Acceptability is required in order to create the best possible translation that fits to the target reader. Similar to readability, acceptability focuses less on accuracy and focuses more on dynamic equivalence.

Skopos Theory
Nord (1991) sees Skopos theory as an intentional translation that must be judged to see how well the text has fulfilled its role in the target language situation. Baker (2001) further explained that skopos is a new perspective in translation study that reflects a more functional and socio-cultural orientation in a translation, where the source text exists merely to provide the main information to produce the target text. This is in line with Nord (1997), that deters the function of the source text to emphasis on the target translation, and that the translation result may differ -slightly or drastically-from the source text. Skopos is essentially a communicative translation method but with added function of goal-and action-oriented purposes. In addition, skopos is necessary to produce a highly practical application in the target language.
Skopos is commonly found in the texts that are related to culture. This include: advertisements, novels and stories, movie subtitles, songs, and poems. These texts can be translated just as well using the translation techniques and methods using Molina & Albir's theories (2002), but skopos always puts heavy emphasis on the target text; whereas Molina & Albir's methods and techniques always put more emphasis on the source text.

METHODS
To answer the research question, I used descriptive qualitative research. The questionnaire was taken by the users of Xiaomi Redmi Note 4, both by the anonymous people on the internet and in real life, and all with their consents. It was conducted with a minimum a total 25 takers combined: three professional translators and 23 end-users. The questionnaire rubrics were differentiated for the translators and the end-users. For the translator, three four rubrics were used: Goff-Kfouri's and Nababan's rubric for translation accuracy, and Nababan's rubric for translation redability and translation acceptability. For the end-users, only two rubrics were used: Nababan's rubric for redability and acceptability.
For assessing accuracy, both Goff-Kfouri's and Nababan's rubrics provide different point of views for the assessments. Goff-Kfouri's rubric provides in-depth analysis on each textual elements of the translation, with focuses on Fluency/Flow, Grammar, Terminology, General Content, and Mechanics. The ratings from Goff-Kfouri's rubric were not summarized, rather they were assessed individually and for each three translators. For the Nababan's rubric on assessing accuracy, it gave general view on the translation accuracy, as well as personal comments on each of the questionnaire items.
There are several reasons why I use two instruments that look radically different in terms of analysis. The first is because Goff-Kfouri's rubric does not regard translation technique as important as the other factors such as grammar and general context, which means translation equivalence is not very important, and it is inline with the Skopos theme in my research. It also offers more detailed marks on each translation aspect. Nababan's rubric on the other hand, deals more with equivalence between source and target text. It also signifies a more general approach in assessing the accuracy. In conclusion, both specific and general views are equally important in assessing translation accuracy For assessing translation readability and acceptability, I used Nababan's instrument to assess the readability directly from the questionnaire takers. Readability analysis usually takes quantitative forms, with up to ten types of calculation. The calculation result from all or part of them would then be compared with average reading competence in certain regions or countries. The problem with using such calculation in translation is that, it does not really show how good or how bad is a translation from a subjective point of viewespecially by the actual readers-and the result cannot be explained and further elaborated in accordance to my purpose. Nordquist (2018) shares his idea about readability formulas, that they do not serve any purpose at all in actually providing information regarding a translation readability among the true target readers. He also insists that readability is best being assessed by the true target readers themselves. Certain formulas such as Felsch' or SMOG's, require a minimum of 100 word-count to be able to be assessed, and texts found from cellphones are mostly very short, thus will not produce any result at all. This also applies to acceptability analysis, where the instrument is similar to readability's.
In assessing readability and acceptability, the assessment was taken from the answers by the end-user directly. If end-users gave their rating of more than 62.5% of average (rated 3.4~5 for readability, and 1.0~3.4 for acceptability) then it implied positive understanding, thus the translation has good readability. If the respondents show negative understanding (rated 1.0~3.4 for readability, and 3.4~5 for acceptability), then it would be concluded that the original translation has readability error. Negative Understanding is where the questionnaire takers do not understand the context of the original translation provided in the questionnaire, whereas Positive Understanding means that the respondents understand the meaning implied. The positive and negative understanding are taken from each word or sentence by face-value.
The original translations are provided along with the modified translation of the same context in random orders in each question. The modified translations act as the distractor and a comparison for the original translation. In addition, if the average readability of text A is higher than average readability of B, the conclusion is that there is no readability error; and if vice versa, there is a readability error.

RESULTS AND DISCUSSION
In this part, I presented the result of the translation accuracy, readability and acceptability, in this respective order, after assessing the texts from Android Xiaomi Redmi Note 4 smartphone.

Translation Accuracy
The assessment of translation accuracy is only conducted for three professional translators with two instruments each; one by Goff Kfouri and the other by Nababan. The result of the accuracy assessment is determined for each number group (1a and 1b, 2a and 2b, etc). For Goff-Kfouri's rubric, result were taken from the average ratings of each number for each criterion, without the ‗Mechanic' criterion. From the results by the three translators, 4 criteria from Goff Kfouri's rubric showed 2.8 for original translation and 3.6 for modified translation; whereas the ‗mechanics' rating was 3.3 for original translation and 3.7 for modified translation. The details can be seen below, where -#a‖ indicates original and -#b‖ indicates modified translation. In assessing the translation accuracy using Goff Kfouri's rubric, not only that the question items were assessed at face value, but I also added the backtranslations for comparing the actual formal equivalence with the original translation, and with the accuracy ratings the translators gave. The analysis on this part of the analysis was actually done one-item by oneitem, and there were no actual ‗overall analysis' except that from the average ratings, the modified translation was found to have better accuracy.
The individual analyses resulted in several interesting findings. First of all, the translators did not actually assess the pure accuracy of the texts, but they actually regarded the accuracy in relation with mostly readability. The first instance of this occurred by the first translator, on question item number -3‖, where the rating for ‗Terminology' criterion spiked from 2 in -3a‖ to 4 in -3b‖-despite the original translation was a faithful, formal translation.
There has yet any theory to mention this yet, so I call this a -paradox of accuracy‖.
Another interesting phenomenon I found out was that, in certain instances, the translators think that some commonly used terms in Android translation, or technology-related translation, were actually less accurate (or probably less readable) than they actually worth. This was first seen in question item -5a‖, regarding the translation of -uninstall‖. The weakness of Goff Kfouri's rubric is that it does not provide an open-ended space for the assessors. Comments are especially important for accuracy, because I need the reason why the translators put their ratings as such. The instances above could not be explained with mere numbers. This is where Nababan's rubric comes in.
The second translation accuracy assessment uses the instrument by Nababan (2012in Hartono, 2017. This instrument is simpler but gives a space for open-ended comments by the translators. The scale for rating for Nababan's instrument are longer and slightly more complicated, but envelops all aspects at once; contrary to Goff-Kfouri that details each and every criterion with simpler explanation for the ratings. The ratings in Nababan's rubric are reversed than Goff Kfouri's, hence 5 means the worst and 1 means the best. The analysis for this rubric was individual. Each question item (or sometimes a couple) was analyzed individually, while sometimes comparing the results with their respective Goff Kfouri's analysis and/or the backtranslations. Just like the Goff-Kfouri's analysis, there were also several interesting bits to take notes on in this analysis. The first instance among others was found in the first translator's question item number -5‖. In this item analysis, the first translator claimed that he has never heard of -bongkar‖ to translate -uninstall‖, and he put both -5a & 5b‖ in not very accurate rating. His comment on -5b‖ was that he agreed to translate -install‖ to -instal‖ but disagree to translate -uninstall‖ to -uninstal‖. It is okay to translate -install -to ‖instal‖ while -uninstall‖ to -bongkar/uninstal‖. If this looks inconsistentboth are two terms of the exact same context, but only antonyms-it is simply how it is written in English language. English uses -un-‖ prefix to show -not-‖, but Indonesian language.
Another example for Goff Kfouri's analysis was taken from the item number -4‖. For this number, all three translators gave better rating for the modified translation than the original, however in fact the original translation is perfectly equivalent with the source text. Despite that, all three translators think that the accuracy of the modified translation -that was less formally equivalent than the original translation. The difference between them was the grammatical element, where the original translation applied the exact same grammatical and structural rules as the English source. The conclusion for this item was that the translators might have thought of the readability along with the accuracy; thus accuracy without readability was deemed incomplete.
In the third example, taken from question item number -1‖, it was an example of how the original intended meaning affects the translation. Text in android are usually short and does not have any cultural backgrounds and/or appropriations. How the texts relate to each other was from where the text was found at, and the associated text related to them. In inspecting -1a‖ against the intended meaning and the location of this text, this text was found under the -Security‖ menu, inside the general -Settings‖. In this premise, the end-users actually did not need the additional text. The end-users already knew what the exact function of this text and need no additional help, hence the original translation served its purpose well. However, there were some chances where a user finds this text in other situations, e.g. from the ‗search' function, or from the ‗help' page within the phone, or simply reading the one-off text somewhere else. In the latter case, additional information would help the reader understand the immediate context without having to research too much of the related background information. In conclusion for question item number 2, it still has a low accuracy rating. Hence, the accuracy with slight relation to readability of this text number -1‖ was less satisfactory that the modified translation scored a better rating.
For ‗mechanics' part, all five question items are analyzed as a whole. Mechanics is how the translation is done, how much is translated-be it contains omission or addition-and simply whether it fits or not. It is almost related to acceptability. The average mark for the whole question items are 3.5; detailed at 4 by the first translator assessor, 3 by the second, and 3.5 by the third. The first and second translators only filled homogenous numbers in all the question items; the first translator assessor wrote 4, and the second writes 3. These translators were assumedly did not actually understand the meaning and intention of Mechanics. For the third translator, the average of -5b‖ beats -5a‖ by 1 rating, that changed from 3 to 4. In conclusion on Mechanics, is currently unmeasurable and unable to be analyzed due to the other two translators did not fully understand how to rate them. In conclusion, this ‗Mechanics' analysis would have deemed as better to be ignored. From Nababan's rubric, the first interesting instance was found from the first translators' assessment on question item number -3‖. The translator neither agreed or disagreed with the shown terminology for -bar‖, translated to -bar‖ in -3a‖ and -bilah‖ in -3b‖, and put the rating 3 for both. He gave an alternative, that is -panel‖. In this Nababan's rubric, rating 3 means slight accuracy problem, and he commented on the terminology only. In relation to Goff-Kfouri's rubric, where the first translator gave rating 2 in -3a‖ to 4 in -3b‖, with the original translation was a actually faithful, formal translation, it was proven by his comment where both -bar‖ (the borrowed translation) and -bilah‖ were thought to be not good. The second translator on further commented on item -3‖ for a new translation which is -panel‖. Updated from the translator's term preference, the translation becomes -Notifikasi dan panel status‖.
The second instance was taken from the first translator's question item number -5‖. Here the translator commented on -5a‖, where he claimed to never have heard of -bongkar‖ (lit: disassamble) being used to translate -uninstall‖. In -5b‖ he said that -uninstal‖ (with one ‗L') was not a good choice to translate -uninstall‖, however he also said that he agreed on using -instal‖ (one ‗L') to translate -install‖. In the respective Goff-Kfouri's analysis for this question item, I concluded that -bongkar‖ was used in more practical fields, whereas -uninstal‖ was simply the transliteration of the source. Within the Skopos theory, although equivalence did not matter, consistency was still important.
Consistency is when a term was used in consistent way throughout the document according to its correct context. In this example -instal‖ was already correctly used within the scope of applications, hence it was correct and consistent.
Another interesting bit by Nababan's instrument to assess translation accuracy was seen at the question item number 1, where -1a‖ got a 5 and -1b‖ got a 1. This absolute drastic change is accompanied with a comment that implied alleged accusation for machine translation, because -for‖ was directly translated to -untuk‖. This comment by the second translator was related to grammar and structure, as -for‖ was actually correctly translated. However, it is deemed inaccurate and caused another paradox of accuracy. Both backtranslations of -1a‖ have proven thiswhere both were formally equivalent to the source. Same as the structure, as was is also proven to be exactly the same as the source text. In conclusion for this item, this result further strengthened the Skopos theory, that accuracy does not always mean good translation, and that the readers' perception is much more important.
In conclusion for translation accuracy by both instuments, the accuracy of the original translation is worse than the modified translation. In Goff-Kfouri's non-averaged rating results, the majority of the ratings showed better ratings for the modified translation. This was further proven by Nababan's rubric, where the average ratings of the original translation by the first, second, and third translator are 2.4, 2, and 2.2 respectively; whereas for the modified translation, the ratings are 3.8, 3.6, and 3.4 respectively, which indicate better accuracy.

Translation Readability
The translation readability was assessed by the three translators and the end-users. In readability analysis, the lower the number is, the worse the translation readability is, with the range of 1 to 5. From the three translators, the total average is 2.3, tallied at 2.3 by the first translator, 2.3 by the second translator, and 2.2 by the third translator. Meanwhile, the total average for the modified translation falls at 3.8, with 3.8 by the first translator, 4 by the second translator, and 3.6 by the third translator. Thus the original translation had negative understanding, while the modified translation had positive understanding. For the result by the 23 end-users, the average rating for the original translation by the end-users is 2.4, and it indicates a negative understanding. The average rating for the modified translation is 3.8, which indicates a positive understanding. Thus it is concluded that the modified translation has better readability than the original translation.

Translation Acceptability
The translation acceptability assessment took the same approach as the readability counterpart. In acceptability analysis, the lower the number, the better the translation acceptability is. For the translation acceptability assessment by the three translators is that the acceptability of the original translation is lower than the modified translation. The rating for original translation is 2.5, and the modified translation, the rating is 3.6. In this analysis, the original translation is unacceptable whereas the modified translation is adequately acceptable. Meanwhile, the translation acceptability result by the end-users showed that the original translation is very unacceptable with 2.2 rating, whereas the modified translation is fairly acceptable with 3.7 rating. In conclusion, the modified translation has better acceptability than the original translation.

CONCLUSION
In conclusion by all the analysis, the quality of the translation that is accuracy, readability and acceptability of Xiaomi Redmi Note 4 is less than satisfactory, as proven that the modified translation always gets higher overall rating than the original translations.