Lexical Errors Made by Instagram Machine Translation in Translating the Account of “CNN Indonesia” News Article

. This study aims to depict the types of errors made by Instagram Machine Translation and to find out the most dominant types of lexical errors made by Instagram Machine Translation on ‘CNN Indonesia’ Instagram account. The design of this research was qualitative design. The data were in words, phrases and sentences that contained lexical errors made by Instagram Machine Translation on the “CNNIndonesia” Instagram account. The data were taken by running an Instagram in one account of various captions related to the lexical errors of the study object. The data were collected through stages: finding out and determining, classifying and separating the words, phrases and sentences that contained lexical errors made Instagram Machine Translation on the “CNN Indonesia” Instagram account. The techniques of analysis data researcher translated the captions using Instagram machine translation and then the translation result is compared to the source language. The next step is to examine the lexical errors produced by Instagram machine translation. The research result shows that the types of lexical errors made by Instagram Machine Translation on the “CNN Indonesia” Instagram account based on the error categories theory by Vilar et al founds are: 4 missing errors, 10 incorrect words and 8 unknown words. All errors indicated that Instagram machine translation could not represent the target language in the “CNN Indonesia” Instagram account. The users of Instagram need to filter every translation that is translated by Instagram machine translation before receiving it as information .


Introduction
The concept of translation introduces to communicate messages from one language to another.
Translation is an activity that aims to convey the meaning or meanings of a given linguistic discourse from one language to another.Translation may be described in phrases of sameness of meaning across languages.Translation is a challenging task in transferring meaning from a source language (SL) to a target language (TL).It's concluded for this sense because an irresponsible translation system will result in a misunderstanding of the message located inside the supply language within the target language.The equivalence of a translation must be expressed in an appropriate way in SL to TL so that readers can enjoy the translation and forget for a moment that they are reading is only a translation [1].
According to [2] when translating, it is important to understand the meaning of the source text in order to create an accurate translation in the target text so that meaning is translated in terms of grammar, style and sound.It is very important to understand the meaning of the source language when we translating, by understanding the meaning of a source language the translating will reach the target language according to the structure, writing style, and sound of the writing.And we need to be able to understand or be proficient in the two languages we are going to translate so that the translation does not look awkward [2].
Nowadays in the world of translation it is different in translating the text.People are getting used to using machine translation.But still, machine translation still needs to be check by human rectification.This is to avoid the translation error found in the translation result.Machine translation means automatic translation.Machine translation is designed to translate text from one language (source language) to another (target language) without human help.Machine Translation offers a machine that interprets textual content from the source language to the target language.The interpretation expresses the same meaning as inside the source language [2].
The "Incorrect Words" errors are the most general type of error.When the system is unable to find the correct translation of a given the word.We distinguish five subcategories here.The incorrect word in the first example distorts the meaning of the sentence.In this case, we could further distinguish two additional subclasses: when the system selects an incorrect translation and when the system is unable to disambiguate the correct meaning of a source word in a given context, though the distinction is certainly hazy.The following subcategory of "Incorrect Words" errors occurred when the system was unable to produce the correct form of a word, even though the translation of the base form was correct.It is especially important for inflected languages, where the high variability of open word classes makes machine translation difficult.
Extra words in the generated sentence cause another type of error.This type of error was introduced when researching the translation of spoken language input, as artifacts of spoken language may produce additional words in the generated sentence.
The last two classes are of lesser importance.The first is style errors refer to a poor choice of words when translating a sentence, but the meaning is preserved even if it is not entirely correct.
A common example is the repeated use of a word in a close context.A translator would choose a synonym and avoid word repetition in this case.The second is about idiomatic expressions that the system does not recognize and attempts to translate as normal text.Normally, these expressions cannot be translated in this manner, resulting in additional translation errors.
Unknown words can also cause errors.In this case, we can differentiate between truly unknown words (or stems) and unseen forms of known stems.A variation of this category is particularly important for the Chinese-English language pair.The majority of European languages, or even languages with the same alphabet, can be "translated" simply by copying the input word to the generated sentence, with no further processing.
Lastly, there may be punctuation errors, but for the current machine translation output quality, these are minor annoyances for languages with no fixed punctuation rules and are not taken into account further in this work.Of course, the error types defined in this manner are not mutually exclusive.In fact, it is not uncommon for one type of error to result in another.A bad word translation, for example, can result in a bad ordering of the words in the generated sentence.
Then, according error classification based on [4], actually, there are only three categories are related to the type of lexical error.As in [6], many of the errors made were as a result of the student's lack of understanding in differentiating the word class.There are missing words, incorrect words and unknown words.For instance, to reveal the adjective form, the students did not add the suffix to the noun.
Instagram is one application that is booming and is used by many people in the world.Quoting

"An Analysis of Grammatical Errors Of Using Google Translate From Indonesia To English In
Writing Undergraduate Thesis Abstract Among The Students' English Department Of Iain Metro In The Academic Year 2016/2017" by (Kurniasih, 2017) in this thesis she focused on grammatical error in translation.This study applied Miles and Huberman.The research result show that the student using Google translate in translating the abstracts and show results of a finding translate by Google Translate is not accurate in English [7].
The similarity with my research is the topic about machine translation.The differences are the focuses in Kurniasih's research is on grammatical error while this researcher focuses on lexical error and in Kurniasih's used [8] while this research refer use [4].Previous study contributing to enhances the writer knowledge about Miles and Huberman theory and shows that Google Translate is still not accurate.
"An Error Types Analysis on YouTube Indonesian-English Auto-Translation in Kok Bisa?
Channel" by [9] in this article author investigates the error types that commonly occur in the translation produced by YouTube auto-translate.This research uses error classifications from This study used Morgan's sample selection table.The errors were categorized based on the classification of error types developed by [11].The results of the study revealed that the register category was the most frequent error area.
The similarity with this research is the topic about machine translation.The differences are the focuses in Jahanshahi's analysis of the type and frequency of the errors occurring in the English translation of Islamic texts by Iranian translators and analyze the possible cause of the errors while this researcher focus on lexical error made by Instagram Machine Translation and in Jahanshahi's research Morgan's sample selection table, while researcher refer use [4].Previous study contributing to enhances the writer knowledge about Morgan's sample selection table and revealed that errors occurring in the English translation of Islamic texts by Iranian translators and the possible cause of the errors revealed that the register category was the most frequent error area.
"Lexical Errors Produced By Instagram Machine Translation" by [12] in this thesis she focuses on lexical error in translation.This study applied theory of [4].The research result shows that Instagram machine translation produced so many errors and shows the weakness of machine translation to represent the genuine language.
The similarities with my research are the topic about machine translation, and then the thesis focuses on lexical error in translation and use theory of Vilar [4].The difference between Susanti's and mine is only Instagram account.But the application we use is the same.Previous study contributing to enhances the writer knowledge more about Vilar theory and shows that Instagram translation is not represent the genuine language.

Discussion
. Table 11 Frequency of types of Instagram Machine Translation From the table above, it can be seen that types of IMT errors based on Vilar et al (2006) related to lexical categories are missing words 18,18%, incorrect words 45,45%, and unknown words 36,36%.The incorrect words shows the highest percentage because incorrect word happened when the system or machine translation unable to find the correct translation in the translation result.

Conclusion
According to the analysis and findings in the previous chapter, it can be concluded that the ten data from the captions on "CNNIndonesia" Instagram account contain three types of errors in the lexical category.Those errors are missing words, incorrect words, and unknown words.But, three types of error are not always found in every data.Incorrect word and unknown word become the highest frequent error found in ten data on the captions of "CNNIndonesia" Instagram account.In general, the error that is often encountered in IMT is translating the trade name or the name of the Institute literally.Thus, there is an unnecessary translation process.In addition, all errors indicated that Instagram machine translation cannot represent the target language in the "CNNInndonesia" Instagram account.So, the users of Instagram need to filter every translation that translated by Instagram machine translation before receiving it as the information.

REFERENCES
photos posted on Instagram.This feature will make it easier for many people to know the translation of captions in other languages.

[ 4 ]
. The result shows that the most frequent error types are wrong lexical choice, bad word form, missing auxiliary word, short range word level word order and extra word.The other error types rarely occur in the translation.The similarities with my research are the topic about machine translation and uses error classifications from [4].The difference are the focuses in Laksana's research is investigates the error types that commonly occur in the translation produced by YouTube auto-translate while researcher focuses on lexical error produced by Instagram Machine Translation.Previous study contributing to enhances the writer knowledge about Vilar theory and shows that YouTube auto translates is still made error."Error Analysis of English Translation of Islamic Texts by Iranian Translators" by [10] in this journal author analysis of the type and frequency of the errors occurring in the English translation of Islamic texts by Iranian translators and analyze the possible cause of the errors.

Table 1
Classification of Data 1

Table 2
Classification of Data 2

Table 3
Classification of Data 3

Table 4
Classification of Data 4

Table 5
Classification of Data 5

Table 6
Classification of Data 6

Table 7
Classification of Data 7

Table 10
Classification of Data 10