This chapter introduces the rationale of the research along with the participants, materials, and instruments used as well as the procedure of data collection, ending with a section of data analysis.
Research Design
This research used both quantitative and qualitative methods to analyze the differences of written translation speeds between translators and interpreters. The time spent on each phase of the translation process are measured in seconds by electronic timers.
The participants are asked to translate two texts of the same difficulty and the written translation process is divided into two phases, draft and revision, which is based on Jakobsen’s (2002) suggested model. Jakobsen divided the written translation process into three phases: orientation, drafting, and revision; in this study, the process of written translation consists of only two phases, drafting phase already including the orientation phase.
The design of this research aimed to first verify if interpreters may do written translation faster than translators did as claimed by some previous studies, where the researchers asked some translators and interpreters to manually translate and sight translate the same texts, and then compared their written translations and transcribed sight translations (Dragsted &
Hansen, 2009; Dragsted et al., 2011).
Next, this study attempted to deliberately speed up the process of written translation by reducing possible interruption in the draft phase of the written translation process.
Specifically, the experiment tried to elicit smooth delivery in the written translation process by avoiding correction keystrokes. Certain keys and the mouse, which could be use to correct translation, are disabled during the draft phase of the second text to see if such limitation imposed brought about changes in speed and translation strategies adopted.
Adjustments in the Present Study
Research methods chosen in the present study. This study used retrospective verbal
reporting as part of the interview. The researcher has avoided introspective reporting, such as think-aloud protocols (TAPs), because the differences between spoken and written modalities lie at the core of the present study and hence the possible interference between the two should be avoided. As Toury (1995), a written translation process that involves introspective reporting might, after all, produces “spoken-written” translations instead of just “written”ones (p. 236). Along with the retrospective interviews, the researcher also used questionnaires, keystroke logging and compared the participants’ target texts. In other words, like recent researchers in the same area of study, the researcher has chosen triangulation, combining different methods to look into the phenomena that occurred in the translation process.
Improvements made in the present study. The current study, based on the inspiration
drawn from previous studies (Dragsted & Hansen, 2009; Dragsted et al., 2011), is improved in several aspects for better validity of its research results.Quantitative improvements. The current study made the following quantitative
improvements:1. Sample size: This study aims to build upon the previous studies (Dragsted &
Hansen, 2009; Dragsted et al., 2011) and refine their research methodologies in terms of the sampling size, language direction, and variables controlled.
Previous studies comparing translators’ and interpreters’ performances used a small sample size of six to 14 experiment subjects. Such a small number of participants hardly yield quantitatively meaning data. Hence, this study recruited 32 subjects (or 33, including the participant for the pilot test) to produce
quantitative analysis.
2. Length of source text: These previous studies used very short texts.
3. Precision of speed & quality of translation: In the experiment conducted as part of the present study, time is measured in seconds by electronic timers and the participants’ translation products are rated based on a 100-point scale by experienced professional translators and editors.
Qualitative improvements. The current study made the following qualitative
improvements:1. Retrospective interviews: The verbal reporting is done after the translation experiment is finished, the researcher asking the participant specific questions and present his/her drafts/revisions. Doing so, the researcher avoided possible interference of the oral modality that might occur during introspective verbal reporting such as TAPs.
2.
Control of the texts translated: The vocabulary complexity and readability, an translation difficulties of the source texts are carefully controlled to avoid possible interference.3.
Proximity among different tasks: The translation of the two source texts and the interview are done at once. Such proximity eliminated possible influence from time. For instance, in the study mentioned beofre, the six participants translated the texts three years after they have interpreted them.4.
Language combination and direction: The two previously mentioned studies asked their subjects to translate into their B language, in this case from Danish into English, which might involved more factors. (Dragsted) The present study asked all the participants to translate into their native language, Mandarin Chinese. Translating into one’s native language, especially when from a sourcetext that pose no difficulty in comprehension, gives the translator a chance to choose from all the possible corresponding translations because these are quite available in his/her native language according to Gile’s Gravitational Model. “It is not unreasonable to assume that on average LC production availability is higher in one’s native language and thus allows richer and perhaps more idiomatic speech production than in one’s B language” (Gile, 2009, p. 237).
“In translation, language availability is not a strong detreminant of performance, but editorial quality, and in particular the stylistic quality of the product, are important. For this reason, even in LSPs, it makes sense to foster translation into one’s A language. On the other hand, if market conditions impose working into one’s B language, this can be done both from one’s A language and from one’s C language. In interpreting, assuming that overall availability is lower in a B language than in an A language and that it is even lower in a C language, working from a C into a B entails higher risks of saturation and is avoided (Gile, 2009, p. 238).
In addition, the present study analyzed translations from English into Chinese, two language from different families. Because these two languages are of very different linguistic features and cultures, analyses yielded more interesting results.
For instance, the linear translation analysis is of more relevance because the word orders are in general different in these two languages.
5.
Comparability: The present study compared the translators’ written translations with the interpreters’ written translations of the same texts. Such comparison can more genuinely reflect the differences between translators’ and interpreters’performances.
The previous studies sometimes reached conclusions based on comparing different things. Specifically, interpreters’ interpretation transcribed by hand or
speech-to-text technology is compared to translators’ written translation. One wonder if comparing different produces can yield any meaningful comparison.
Hence, this study asked translators and interpreters to performance exactly the same tasks—they did written translation and written translation only.
Despite the potential for speech-to-text technology to replace man-made transcription, such technology is still not very useful for recognizing languages like Chinese for they have a great number of homophones, which might lead to constant error in speech recognition.
Participants, Materials, and Instruments
The researcher recruited a total of 32 experiment subjects, namely 10 translators and 22 interpreters, who have trained for at least one year in the Graduate Institute of Translation and Interpretation, National Taiwan Normal University. Most of them have hands-on experience in written translation prior to the experiment.
The participants are instructed to translate the two texts in simulated context: The texts, which are about general topics and are meant for common readers, are part of a book left untranslated and now asked by the editor to translate as soon as possible. The key instruments for data collection are Translog—a software program that recorded keystrokes—in the drafting phase and Microsoft Word in the revising phase.
Participants. The 32 participants included 20 current students, one dropout, and five
graduates from the master’s program as well as six current students from the PhD program.Because they all have passed the program’s highly competitive entrance examination (Chang, 2011) and obtained at least one year’s professional training along with varying degrees of hands-on experience, they are referred to as translators and interpreters in this study. Nine of them are male and 23 female. Most of them majored in English or related fields before
entering the graduate program at GITI, NTNU (see Appendix D).
Materials. The two English source texts, Text A and Text B, are excerpts from two
books which the researcher personally translated for publishers.1 Both books, one of which about entrepreneurship and the other about critical thinking, involved no technical content and are meant for general readers. Text A and Text B consisted of 242 and 243 English words respectively.Most of the words used in the two texts came from Oxford 3000TM keywords2, a list of the most frequent words in the English language (Oxford University Press 2011). According to Oxford Dictionary’s Online Text Checker3, Text A and Text B fell under the category of high intermediate text because 90-95% of the words in each text is Oxford 3000 keywords (see Table 7). The participants, whose English skills are good enough to train as professional translators and interpreters, should not need to consult dictionaries or online resources, which they have no access to during the experiment. Exclusion of such access decreased the number of variables, such as online research skills, which might affect the experiment results.
Table 7. Vocabulary Difficulty Check against Oxford 3000 Keywords
Text A Text B
*Percentage of Words belonged to Oxford 3000TM Keywords
94% 95%
To control for similar translation difficulty, each excerpt has been slightly modified to almost the same length and readability (see Table 8). Specifically, this study used two very
1 Book A: It's a Jungle in There: Inspiring Lessons, Hard-‐Won Insights, and Other Acts of Entrepreneurial Daring by Steven Schussler; Book B: How to be a Brilliant Thinker: Exercise Your Mind and Find Creative Solutions by Paul Sloane.
2 The list of Oxford 3000 keywords and its purpose:
http://oaadonline.oxfordlearnersdictionaries.com/oxford3000/
3 The Oxford Text Checker which assesses the percentage of words that belong to the Oxford 3000 keywords:
http://oaadonline.oxfordlearnersdictionaries.com/oxford3000/oxford_3000_profiler.html
common readability formulas, Flesch-Kincaid Reading Ease and Flesch-Kincaid Grade Level (Burke & Greenberg 2010).
Table 8. Readability of Source Texts
Text A Text B
Character Count 1,058 1,095
Syllable Count 353 342
*A higher score indicates easier readability; scores usually range between 0 and 100.
**A grade level (based on the USA education system) is equivalent to the number of years of education a person has had. Scores over 22 should generally be taken to mean graduate level text.
Instruments. Translog, a Windows program that could record keystrokes during the
translation process, is used to record how many times certain keys are pressed. This program is rarely utilized in Taiwan for it only ran on the XP Windows platform and could not able to record any Chinese character input, nor could it record any information on character conversion during the input process of most Chinese input methods. This study used it torecord stroke of correction keys, namely Backspace, Deletion, and the mouse.
SPSS 22 (Statistical Package for the Social Sciences), a commonly used statistics software program, is used to statistically analyze the collected data. Basing on the analyses on SPSS’s t tests, the researcher is able to check the quantitative significance of the data, which can be counter-intuitive in some cases.
During the retrospective interview that took place immediately after each experiment, the conversation is recorded using ClearRecord, an iOS app on the researcher’s iPhone 4. The recordings are AAC (Advanced Audio Coding) files with qualities of 16kHz. The recordings are transcribed suing Dozitenggaoji (豆子謄稿機), a Windows program that allow users to fast forward or rewind as they type and listen to the audio file on the same window.
Data Collection Procedure
The researcher himself set up the experiments, collected the participants’ translations, and conducted the interviews. Specifically, the data were collected in the following three stages, one stage immediately followed by another: Pre-Experiment Set-up (Figure 2),
Experiment (
Figure 3), and Post-Experiment Interview (Figure 5).
Figure 2. Pre-Experiment Set-up
Step 1
• The researcher posted a Do-Not-Disturb notice on the entrance of the experiment venue.
Step 2
• The researcher installed the participant's preferred input method on the laptop.Step 3
• The researcher explained the procedures and instructed the participant to sign the consent form.
Figure 3. Experiment
Figure 4. Snapshot of Translog
*Note: this window showed what a participant typed (i.e. his/her translation). The
STEP
1
• WARM-UP 1: The participant typed a specified short Chinese passage on Translog.
STEP 2
• DRAFT 1: The participant translated Text 1 on Translog.
STEP 3
• WARM-UP 2: The participant typed a short Chinese passage on Translog.
NOTE: Backspace/Del/Ins keys disabled; mouse unplugged.
STEP 4
• DRAFT 2: The participants translated Text 2 on Translog.
NOTE: Backspace/Del/Ins keys disabled; mouse unplugged.
STEP 5
• BREAK
STEP 6
• REVISE 1: The participant revised his/her draft translation of Text 1 on Microsoft Word.
STEP 7
• REVISE 2: The participant revised his/her draft translation of Text 2 on Microsoft Word.
original text is presented on paper, which is placed next to the laptop.
In the beginning of the experiment, the particpant was instructed to type a short Chinese passage about the simulated context—“Imagine you have trasnalted and submitted a book a month ago. Now the editor told you that you accidentally left a passage of three paragraphs untranslated, which is the text you are going to translate in this experiment. Because the book has been scheduled to be published soon, the editor wants you to translate this passage and email him back as soon as possible.” In addition, the reseacher reminded the participant that it is the speedy translation s/he should be aiming for and although the quality of translation is important, it only have to be good enough to keep this customer (the editor in this case) coming back for the same service. In other words, the participant was instructed to achieve the fastest possible speed with an accpetable quality.
This warm-up session provided the participant with a chance to familiarize him/herself with the hardware (laptop’s screen, keyboard, etc.) and the interface of Translog (Figure 4).
The passage typed also reminded the participant of the simulated context.
After that, the participant translated Text 1, knowing that this is the drafting phase and s/he will be able to refine his/her draft during the revising phase.
Before the participant translated Text 2, s/he warmed up again to familiarize him/herself with the limitation imposed by the experiment—the three correction keys, i.e. Backspace, Del, Ins, and the mouse are disabled. The participant typed the same short Chinese passage, this time not able to use the aforemetioned correction keys.
After the warm-up session described above, the participant began translating Text 2 under the same constraints.
Then, the participant took a break before revising Text 1. It should be noted that the interfaces used for the drafting phases and the revising phases are different: the former used Translog and the latter Microsoft Word. The only difference between the translation
processes of Text 1 and Text 2 is that the keyboard and mouse constraints are imposed during the drafting phase of Text 2.
The participant continued to revise Text 2’s draft after revising Text 1’s draft.
Figure 5. Post-Experiment Interview
The researcher asked the particpant to compare the experiment and his/her usual way of translation, focusing on those differences that affected his/her speed and quality of translation on the day of the experiment. Specifically, the participant was asked to explain his/her responses and strategies towards the disabling of correction keys and the mouse. the participant was also asked to compare the difficulties of the two texts and their two translations, explaining the differences in detail. Also, the participant was asked if there was anyting similar to interpreting during the whole translation process.
Before the interview concluded, the reseacher summarized the participant’s responses to the very first question—“how do you cope with the disabling of the three correction keys and the mouse?” and check if the participant would like to correct or add anything. At the end of the interview, the researcher thanked the participant and gave him/her a convenience store gift voucher.
Table 9. Main Interview Questions Step 1
• The researcher asked the paricipant about what has happened during the experiment.
Step 2 • The researcher summarized what the paricipant had said and ask if there was anything to add or correct.
Step 3 • The reseacher gave the paricipant a gik voucher and thanked him/
her for taking part in the experiment.
Q1: Are there any differences between today’s exepriment settings and your usual working conditions, in particular those that affected your quality and speed of translation today?
Q2: How did you respond to the limitations imposed during the draft phase of Text 2’s translation? Describe your reaction and strategies in response to the disabling of correction keys.
Q3: Next, rate on a 100-point scale each Text 1n terms of difficulty level of reading comprehension, translation, and revision. In addition, describe the differences between the two texts.
Q4: Let me summarize what you said to Question 1. Is there anything you would like to correct or add?
Data Analysis
The experiments, including the pilot test, are conducted between May 2012 and March 2013.
This study collected four types of data: questionnaires, statistical observations during the written translation process, 64 Chinese drafts and 64 corresponding revisions produced by 32 participants against the same two English source texts, and excerpted transcriptions of the interviews. The 64 drafts were collected as RTF files and the 64 revisions as Microsoft Word files. The temporal information (in seconds) was analyzed via SPSS to find statistical significant comparisons. The content of the drafts and revisions were compared to find differences in translation strategies between the two groups as well as between the two texts.
The analyses were examined against what the participants said during their interviews, which ranged from eight to 60 minutes.