EXPLORING LINEARITY AND TYPES OF SELF-CORRECTIONS IN THE TRANSLATION OF STUDENT TRANSLATORS

ABSTRACT


INTRODUCTION
Since regarded as a field of study, translation studies (TS) -a name first proposed by James S. Holmes in 1972 -has grown and developed as an academic discipline (Munday, 2012, p. 41). The main factor contributing to the growth and development of TS as an autonomous academic discipline is Descriptive Translation Studies (DTS), which has been a dominant paradigm in TS since the 1980s (Scarpa et al., 2009, p. 32). DTS focuses on three research areas, including product-, process-and functionoriented translation (Toury, 1995, p. 10). Translation product, which used to be the focus of interest in previous TS, often highlights different strategies, techniques, methods, or approaches in translating texts. Such studies tend to cause controversial issues as none of the translated texts is said to fulfil the criteria of a good translation for different techniques or methods used in the translation process. These phenomena imply that a good quality of translation can only be achieved when translated by the researchers themselves. Besides, it is quite ridiculous that we argue with a translation product without any knowledge at who is behind the text, the translator.
For this reason, TS has distanced itself from earlier prescriptive approaches, and many recent studies on translation have a distinctly descriptive focus, by describing what translator are actually doing in the translation process, i.e., the focus of translation process research (TPR). TPR studies how translators work in the three phases of translation process: pre-drafting, drafting, and post-drafting (Mossop, 2001, p. 40). Based on the authors' observations on the translations done by students in some universities in Medan, North Sumatra, many of them spent more time in the predrafting and drafting phases, and spent a small amount of time in the post-drafting phase. This indicates that they focused more on how they wrote the draft of their translation, and almost ignored corrections or revisions on their translation draft. This leads to the assumption that they did not know a very important role of post-drafting phase (correction) in the translation process. As the task was an individual assignment, they had to do correction by themselves.
Doing individual correction or correction on the student's own translation is known as self-correction (Malkiel, 2009) or self-revision (Carl et al., 2010;Carl & Kay, 2011;Rosa et al., 2020). Addition, deletion, substitution, or other kinds of changes involving the choice of words, writing style, and grammar are activities observable in self-corrections. The phenomenon of self-corrections has attracted many researchers to conduct research on self-corrections in translation process with different focuses and methodologies. Those previous studies focused on linearity in doing self-corrections (Asadi & Seguinot, 2005;Kourouni, 2012), procedure for doing self-corrections (Robert, 2008;Robert & Brunette, 2014), self-correction categorization (Malkiel, 2009), validity of self-corrections (Ibarrola, 2009); time allocation for activities in doing selfcorrections (Carl & Kay, 2011), online resources management in doing self-corrections (Sofyan et al., 2016;Sofyan & Tarigan, 2017;Yuliantika & Rosa, 2021), self-correctionbased translation model (Rosa et al., 2020). Moreover, those previous studies have proved the essential role of self-corrections in producing good quality translation products.
The previous studies on self-corrections mentioned above suggest further explorations on self-corrections. In particular, from the studies done by Malkiel (2009)  linearity in doing self-corrections, it is assumed that studying both categories and linearity of self-corrections in one study would be useful. Based on the rational above, this study is intended to find out the linearity and types of self-corrections done by student translators in the translation process.

METHODS
This descriptive qualitative study used process-oriented translation as the approach. The participants were the students of Master Degree Program of Linguistics Department at Universitas Sumatera Utara, Indonesia. They took translation studies as their concentrated field, and in this study, they were called student translators. Using a purposive sampling method, three of them were selected as the participants of this study. Following the participant selection criteria suggested by Sofyan et al. (2016), all the participants were (i) native speakers of bahasa Indonesia; (ii) familiar with major web search techniques for translation purposes; (iii) familiar with the type and style of texts; (iv) not having a professional experience in translation practices; and (v) belonging to the same age group.
Besides, there was a conscious effort to control participant-specific variables in a way that it allowed to see how the quality of the translation was influenced by the process undertaken during translation as suggested by Kourouni (2012). The methods chosen were: (i) background questionnaire (to see their translation experience); (ii) test of bahasa Indonesia (Level of L1), where the proficiency level required was 'unggul' with the range score between 550 and 716 (Maryanto, 2001) as the normative score of bahasa Indonesia proficiency for university students; (iii) test of English Proficiency (TOEFL) (Level of L2), where the range PB-TOEFL (Paper Based Test of English as a Foreign Language) score required for the participant was between 475 and 525; and (iv) test of typing speed and computing skills (using TQ Test), where the range of typing speed was between 25 words per minute (wpm) and 35 wpm with 92% accuracy (Logan & Zbrodoff, 1998;Yamaguchi et al., 2013).
Familiarity with topic and style was a variable that needs to be controlled as much as possible. Several methods should be conducted to get a real sample of the text to be translated by the participants. In this study, the text analysis tools used were: (i) SMOG and Flesch-Kinkaid to consider word frequency and sentence-length formulas; and (ii) Flesch-Kincaid Reading Ease, SMOG, Coleman Liau index, and ARI to test the readability of the texts. Using these tools, the first text selected was entitled 'Apple vs. Google Is the Most Important Battle in Tech" (downloaded from http://business.time.com/2012/10/12/why-apple-vs-google-is-the-most-important-battlein-tech/#ixzz2ADYaIRtb), and the second text was entitled 'The Wholesome Hidden Message of Gangnam Style' (downloaded from http://business.time.com/2012/09/24/the-wholesome-hidden-message-gangnamstyle/#ixzz2ADoZKhk6). The comparison between Text 1 and Text 2 in terms of their readability is shown in Table 1 As this study was qualitative, the main instrument was the researchers themselves. In addition, the researchers were assisted by the supporting instruments including screen recording tools (Camtasia Studio 8) and keylogging (Translog-II). These computer applications had been installed in the computers used by the student translators. They were given enough time to learn how to use Translog-II in the translation process. The chosen source text was given in the Translog work sheet available in the program. Then, the translation process was recorded by using Translog-II and Camtasia Studio 8. All of the student translators were allowed to use online resources as their helpers to support them in the translation process. There was no time limitation; in other words, all of the student translators were given a plenty of time until they finished their translation work. The data were analyzed descriptively to explore the linearity and types of self-corrections in the translation of the student translators.

FINDINGS AND DISCUSSION
The results of data analysis concerning the linearity and types of self-corrections done by the student translators collected by using Translog-II and Camtasia Studio 8 are presented in Table 2. In terms of linearity, as displayed in Table 2, the student translators selfcorrected their translation in two out of three categories of linearity: inline and multidirectional or non-linear. Some of them did self-corrections at a line level, i.e., by focusing on line-by-line corrections, but not following the order of the words in each line. Meanwhile, others did self-corrections multi-directionally, i.e., by leaving the first line for the next lines or leaving the first paragraph for the next paragraphs. They selfcorrected a few lines in a certain paragraph, returned to the lines in the first paragraph, or jumped to the second paragraph, and so forth. Meanwhile, none of them started doing self-corrections from the title and proceeded in a linear fashion until the end of the translation process. Besides, while doing self-corrections employing an inline method, the student translators self-corrected their work words by words, phrases by phrases, clauses by clauses, sentences by sentences, and paragraphs by paragraphs. Nevertheless, mostly they did self-corrections after they finished translating one sentence before translating the next sentences. Based on the log, there was only one case in which the student translator (Student A) did self-corrections after completing the translation of the last paragraph in Text 2.
Based on the data presented in Table 1, Student A and Student C applied a multidirectional or non-linear method in doing self-corrections. They did selfcorrections in a zig-zag manner without giving a priority on the order of sentences or paragraphs. They did self-corrections by leaving some parts in the first paragraph to move to the next paragraphs. Then, they returned to the first paragraph to complete the self-corrections. However, translators should be more careful when choosing a nonlinear method in self-corrections for the possibility of missing to self-correct some of their first draft.
Moreover, this finding shows that self-corrections were not only done exclusively in the post-drafting phase, but it is also done simultaneously with the other phases of translation process. This indicates that self-corrections were also done before the texts were completely transferred to the target language (TL). This finding is consistent with the nature of translation process mentioned in the previous literature (Jakobsen, 2002;Mossop, 2001;Yamada, 2009), translation process is a series of translating activities involving three phases: pre-drafting, drafting, and post-drafting. Nevertheless, the results of this study indicate that the order of the aforementioned translation phases should not always be followed, and denying an order of a series of translating activities does not mean a failure in translating. This finding is consistent with the linearity method proposed by Kourouni (2012). Besides, this finding confirms two approaches in the distribution of self-correction activities in the translation process identified by Asadi and Séguinot (2005). The first approach allows translators to do all the translation activities (writing, researching, and revising) in the three phases of translation process, while the second approach allows translators to do self-corrections only after the source text has been completely translated.
In addition, a certain type of linearity chosen by the student translators is related to their decision making in the translation process. The finding shows that the linearity is caused by the familiarity level of the information pattern of the text. This is in line with Darwish (2008) who says that tactical decision is linear when the information pattern is simple and familiar and cursive for the complex and unfamiliar information pattern. Furthermore, the types of self-corrections displayed in Table 2 are further elaborated in Table 3 to show the distribution and frequency of each type of selfcorrections in the translation of the student translators. The findings displayed in Table 3 show that WS is the type of self-corrections most frequently done by the student translators. This finding is in the contrary with Malkiel's (2009) findings where WD appears to be the type of self-corrections most frequently done by the student translators.
This finding shows that all of them were very careful in choosing the right diction for the right context. The important factor that leads to a very careful choice of words is social acceptability, which is related to the naturalness of the translated text. Online dictionaries provide several possible equivalents to the word in the ST; however, translators should pay attention to the context where the word is located. This finding is consistent with Melby and Foster (2010), arguing that it is impossible to ignore the context in translating a text for its relation to the specifications from which a translator works.
Collocation is also one of the elements of dictions that is different in the SL and TL. Some words in the SL collocate with one or two words, but they do not necessarily do in the TL; therefore, translators have to be very careful in choosing which word is collocated with another word in the TL. This finding is consistent with the previous finding that problems of collocation are the most daunting linguistic problems encountered in translation, particularly because of the non-homogeneity of natural languages (Chukwu, 1997). The other frequent type of self-corrections is WD. Deleting words in the translation process basically aims at improving the translation quality; nevertheless, it does not mean that the translated works containing a number of word deletions result in a good translation. Based on the data analysis, word deletion in self-correction was for the naturalness of the translation product. For example, Student C deleted the word 'dari' in her first draft 'beberapa dari parodi' as a translation of the ST phrase 'some of the many parodies'. In bahasa Indonesia, the word 'beberapa' (equivalent with 'some' in English) can only be collocated with 'dari' when followed by a pronoun as in 'beberapa dari mereka' (equivalent with 'some of them'). However, when it is followed by a noun, such collocation as 'beberapa dari parodi' is not common in bahasa Indonesia. This leads to her decision to delete the word 'dari'.
The finding also reveals types of word deletion done by the student translators (see Figure 1). Previous studies only mentioned word deletion as the type of selfcorrections, but never explored types of word deletion (e.g., Malkiel, 2009;Mizón & Diéguez, 1996). This research particularly found types of word deletion that include deleting (i) unnecessary words, (ii) unnecessarily added words, (iii) incomplete words, (iv) repeated words, (v) miscollocation, and (vi) redundancy. Deleting unnecessary words means deleting the word which is not needed in the TT. For example, the word 'untuk' in the phrase 'untuk internet' written in the first draft of Student B was deleted because it did not have any functions in conveying the message of the ST. Instead, it was used because (i) Student B did not want to miss translating any single word of the ST into the TT, and (ii) the structure of the SL interfered the structure of the TL. The phrase of the TT 'have access to the Internet' contains the word 'access' which can be classified either as a verb or as a noun. In her first draft, she decided to categorize the word 'access' as a verb, meaning 'mengakses'. When it is a verb, it belongs to transitive verb which always needs direct object, not the object coming after preposition. In her final draft, she decided to delete the preposition 'untuk' to allow the verb 'mengakses' to be followed by its direct object 'internet'.

Figure 1. Causes and types of deletion
Deleting unnecessarily added words is much influenced by the dialect of the translator, which may not be the standard language of the TL. Translation is a process of transferring the message from the ST into the TT in written form which requires the standard dialect of the TL. For example, the decision of Student A to delete the word 'cara' in the phrase 'dengan cara membeli' of the first draft for the translation of the ST phrase 'by buying' is considered worthy. Writing the word 'cara' in her first draft means that she added the word which had no equivalence in the ST. Adding a word in translation is acceptable as long as it does not influence the meaning and is not against the structure of the TL. However, in this case, the addition of the word is not necessary because the phrase 'dengan membeli' has already been socially acceptable by the TL readers.
Meanwhile, deleting incomplete words is caused by the carelessness of the student translators which can be motivated by a psychological factor. In their first draft, they wrote the word coming earlier in their mind to represent the meaning of the word in the ST. However, before they had completely written that word, another idea was coming; as a result, they deleted the early word which had not been completely written. This fact, again, shows how self-correction is absolutely needed in translating.
Deleting repeated word is also related to psychological state of the student translators. Sometimes, they did not realize that they had written the same word twice. During self-correction, they reread their first draft and found that some of the words had been written twice. Writing the same word twice can also be caused by a pause taken while writing the draft of translation. Usually, a translator may repeat the word that they have written before taking a pause. Redundancy is the leading factor of deleting words during self-correction.
Student A decided to delete the word 'dana' in the phrase 'dana jutaan dolar' considering that its meaning had been contained in the phrase of 'jutaan dolar'. Deleting the word 'dana' does not reduce the meaning of the phrase. The same condition happened when Student C decided to delete the word 'merasa' in the phrase 'merasa sangat geram'. The word 'geram' (equivalent with 'furious') is related to someone's feelings. Therefore, the existence of the word 'merasa' (equivalent with 'feel') is useless because without which the phrase has already had a complete meaning. The finding on types of word deletion suggests that the term 'word deletion' should be revised as deletion also involves other elements of text such as affixes (suffixes) and phrases deletions. Therefore, the authors suggest the term 'deletion (D)' to replace WD. The deletion involving suffixes includes: (i) pasaran > pasar; (ii) intelektualnya > intelektual; (iii) miliknya > milik; (iv) parodinya > parodi; (v) kebodohannya > kebodohan; (vi) gayanya > gaya. From the six data, there were only two kinds of morphemes deleted: '-an' and '-nya'. Suffix '-an' in bahasa Indonesia, based on KBBI (http://kbbi.web.id/-an), is used to form nouns indicating action product (e.g., catatan (note), bacaan (reading); place or location (e.g., pangkalan (station), kubangan (wallow)); equipment (e.g., timbangan (scale), meteran (length indicator)); and manner (e.g., tembakan (shot)). In this context, the word 'pasar (market)' is a noun and it does not need to be reformed into another noun because its meaning has already represented in the word 'pasar'. Meanwhile, suffix '-nya' is used as: (i) a variant of personal pronoun 'dia (he/she)' and (ii) a possessive adjective. In bahasa Indonesia, sometimes suffix 'nya' is used in meaningless word which is caused by the local language interference. In some regions in Indonesia, such as in Medan, people tend to add meaningless '-nya' in their speech, and it has already become the characteristic of Medan dialect of bahasa Indonesia.
The third most frequent type of self-correction done by the student translators is meaning correction. Meaning, undoubtedly, has the most important role in translation as translation is a matter of meaning transfer from SL to TL. Self-correcting the meaning means adjusting the meaning to decide which expression is more acceptable in the TL. Besides, self-correcting the meaning is done coincidently with both WD and WS because the selected word should have been deleted before its meaning is adjusted, and adjusting the meaning means substituting the word with another word.
Another finding related to self-corrections is that the main focus of the student translators' attention is on how the message in the ST can be well transferred to the TT for clarity purpose, i.e., to make the TL readers understand the message. WS, WD, and M are the three most frequent types of self-corrections done by the student translators. Enclosing the irrelevant additional word, keeping using the less relevant word, and keeping less different meaning in translation influence the TL readers' understanding on the original message contained in the ST.
Furthermore, the finding showing that the student translators corrected the spelling 44 times out of the 416 times of self-corrections (see Table 3) is quite surprising as, unlike English, Indonesian spelling and pronunciation are alike -the words are Exploring Linearity and Types of Self-Corrections in the Translation of Student pronounced as they are spelled. The spelling correction is not a matter of insufficient knowledge of the TL's spelling system, but of the carelessness of the students as the TL is their L1. A correction on spelling errors contributes very much on the translation product since different spelling may result in different meaning. Galinskaya et al. (2014) reported that misspellings affect translation results more than other types of errors, and correction on spelling is the most profitable for improving the translation quality. This implies that misspelling should not be taken lightly by translators.
The fifth most frequent type of self-corrections, i.e., word addition (WA), is a new type of self-corrections found in this research. Previous theories did not consider word addition as the type of self-corrections (e.g., Malkiel, 2009). WA can be defined as the addition of words or phrases to the TT phrases or clauses in order to improve the quality of the TT. Besides, WA is also related with other types of self-corrections, i.e., word substitution and meaning correction. During self-corrections, sometimes the substitution involves WA in order to give sense to the new structure (phrase or clause). Similarly, while correcting the meaning, one or more words added to the new phrase or clause.
The finding on word addition during self-correction is much related to how the student translators wanted to make their translation as natural as possible. Nevertheless, sometimes, adding words may lead to meaninglessness. Based on the data analysis, word addition is much related to the TL structure adjustment. English has its complex structure in terms of tenses which is much different from bahasa Indonesia. The student translators found it difficult to find the equivalence of past verbs in bahasa Indonesia; consequently, they tend to add the word (aspect) 'telah' to show the meaning of the past verb. Another addition is found in self-correcting the meaning of a sentence containing adjective or relative clause. For example, the addition of the word 'yang' to the verb 'diidolakan' causes different meaning of the structure. The verb 'diidolakan' which used to be the predicate, one of the main elements of a sentence, has become the adjective that has secondary role in a sentence. The addition of the word 'yang' also causes what is known as 'clause reduction', i.e., the clause serves as a phrase.
However, as noted previously, word addition may cause meaninglessness. The decision of Student C to add the word 'dana' in the second draft 'dana jutaan dolar' is somehow misleading. This word addition causes redundancy because it gives additional meaning to the structure or the phrase that already has a complete meaning. Another example, Student B added the word 'paling' (equivalent with superlative degree in English) to the word 'top'. In fact, the word 'top' itself in English has already contained superlative meaning, so the word addition causes redundancy in the phrase structure of the translated text. Another finding, capitalization, despite its low frequency, should not also be taken lightly by the student translators. As technology develops, the incorrect use of capitalization at the beginning of a sentence is easy to recognize because the computer program, nowadays, can automatically identify it and provide the correction. However, they must have a good knowledge of capitalization as it may appear in the middle of a sentence. Besides, they also have to be careful in capitalizing the words of a title because function words -articles, prepositions, conjunctions -are not capitalized.
The small frequency of self-correction on return (R) shows that they had only two drafts of translation for most of the self-corrections they did. Meanwhile, return correction indicates that the translator has three drafts of translation. The example from the log done by Student A (marah > geram > marah) is an example of how selfcorrection on return is done.
The finding related to the different frequency of self-corrections done by Student C was surprising. While Student A and Student B did more self-corrections in translating Text 2, Student C did more in translating Text 1. She did 81 times of selfcorrections on Text 1, but only did 69 times on Text 2. According to Flesch-Kincaid readability test, the readability test designed to indicate how difficult a reading passage in English is to understand (cf. Kincaid et al., 1983), Text 2 was more difficult than Text 1. The Flesch Kincaid Grade level of Text 1 was 9.87, while the Flesch Kincaid Grade level of Text 2 was 12.25; this means that she should have done more self-corrections on Text 2. This finding indicated that there should be another factor that influenced the difficulty on translating a text, i.e., familiarity with the topic of the text. Based on the observation done by the researcher while Student C was translating Text 2, she looked more comfortable doing the translation process. This finding was strengthened by the result of the retrospective questionnaire in which she wrote that she did not have too many problems finding the equivalence of the terminology and vocabulary used in Text 2.
In addition, the smaller number of self-corrections done by Student C in Text 2 was caused by her weakness in online resources management. Based on the log, she spent longer time (01.26 hours, compared to 01.23 hours in finishing the translation of Text 1) and longer duration of taking pauses (55 minutes, compared to 43 minutes of pause duration in Text 1) in finishing her translation on Text 2. The longer pauses taken in translating Text 2 should have caused more frequent self-corrections. However, the screen recorded that she visited several websites which were not helpful in doing selfcorrections. She visited Youtube and searched for the songs by the "Gangnam Style" group. Then she played the music and enjoyed it. This made her unable to perform her best efforts in doing self-corrections on Text 2. This finding shows that online resources could become the helper when managed carefully; on the other hand, online resources could also disturb the translators to work properly when it is managed badly.
It should be understood that identifying self-correction means revealing what happened during the translation process. Self-correction is closely related to the translator's decision making whether to use a certain expression instead of others, to delete and or to substitute, etc. Using Translog makes it possible to identify any selfcorrections done by the translators.

CONCLUSIONS
The finding related to the self-corrections describes as well as emphasizes an essential role of self-corrections in the translation process. It can be said there will never be a quality translated work without having been self-corrected. Spending more time on selfcorrection does not mean wasting more time; instead, it can improve the quality of the translated text. Besides, doing self-corrections also makes translators aware of being careful in completing their translation tasks as a small mistake, such as misspelling/mistyping, can influence the quality of their translation.
Doing self-corrections must not always be in line with the prescribed phases of translation process because it can be done simultaneously with the drafting process. Self-Exploring Linearity and Types of Self-Corrections in the Translation of Student corrections at paragraph level is much better, but at word and sentence level is also acceptable. Three methods of doing self-corrections (linear, inline, and multidirection/non-linear) have both advantages and disadvantages. All of them would work well as long as the translators know in which method they can work well.