Dr. Kit Chunyu Wu Ting-Wei Tiffany Translation for Dissemination - PowerPoint PPT Presentation

Dr. Kit Chunyu Wu Ting-Wei Tiffany

 Translation for  Dissemination  Assimilation  Information exchange  Information access  Increasing need and use of machine aids for cost-effective translation

 In the foreseeable future, it is hard for MT to fully replace human translators  How can machines support human translators?  Human-assisted machine translation (HAMT)  Computer-aided translation (CAT)  Editors + additional (computerized) aids:  Bilingual dictionary  Spell checker, grammar checker  Monolingual concordancer  Terminology/ memory database  Web search

 Wide application of CAT tools in industry  Acknowledged benefits to current business environment  Still having call for better user experience

 Why study translation process?  Improvement of translation education  Improvement of user experience translation technology

 Efforts have been dedicated to reformulate the translation process Eg. Comprehension  “Micro - cycle” by Jakobsen (2011)  “Monitor model” by Tirkkonen-Condit (2005) Locate the Locate position current chunk in in TT ST and reread Type and monitor translation

 What subtasks can be identified throughout the translation process?  What is the general distribution of cognitive efforts observed by eye movements during translation process?  How are cognitive efforts distributed among subtasks in translation? Are there any identifiable eye movement patterns within or across each domain/ subtask?  To what extent will the text complexity affect different measures of eye movements during translation process?  Is translation conducted in a sequential fashion as suggested by Gile (2011) or overlapping process can be identified (Hvelplund, 2011)?

 Eyelink 1000 head-mounted tracker  Primitive editor + offline dictionary Target Text Dictionary Source Text

 Task:  1 warmup task + 2 experimental texts of varied complexity  Text 1 > Text 2 (according to the text complexity indicators by Jensen (2011))  All texts has <110 words  Scope of participant:  Translation major students with completion of >1 year of studies  Chinese as the native language

 Two ways of categorizing subtasks  Predefined interest areas  Eye-tracking data + typing events Fixation falls Typing Subtask Type in event ST ✗ ST comprehension ST ✓ Parallel attention (PA) TT ✗ or ✓ TT production Dictionary Dictionary lookup ✗ or ✓

 Overall distribution of cognitive effort  Dwell time  Cognitive workload of different subtasks  Fixation duration  Pupil size  Working style  Cross interest area saccades  Shift probabilities between interest areas  Reading patterns of dictionary

900.00 817.34 Mean total dwell time by Text Type (sec) 800.00 674.52 700.00 600.00 500.00 400.00 300.00 200.00 100.00 0.00 Text 1 Text 2 Figure 6 Mean total dwell time by Text Type (seconds).

379 379 400 400 Mean dwell time by Subtask Type (sec) 362 361 338 350 350 Mean dwell time by AOIs (sec) 298 300 300 250 250 227 195 200 200 150 150 94 94 82 100 82 100 34 50 26 50 0 0 ST TT Dictionary PA ST TT Dictionary Text 1 Text 2 Text 1 Text 2 Mean dwell time by Text Type in different AOIs (seconds). Mean dwell time by Text Type in different subtasks (seconds).

0.6000 Mean dwell time by AOIs (percentage) 0.5354 0.6000 0.5361 Mean dwell time by Subtask Type 0.5000 0.4631 0.5000 0.4635 0.4138 0.4000 0.3646 0.3362 0.4000 (percentage) 0.3000 0.2888 0.3000 0.2000 0.2000 0.1210 0.1210 0.1153 0.1152 0.1000 0.0390 0.1000 0.0419 0.0000 0.0000 ST TT Dictionary ST TT Dictionary PA Text 1 Text 2 Text 1 Text 2 Percentage distributions of dwell time by Text Type in different subtasks. Percentage distributions of dwell time by Text Type in different AOIs.

278 280 275 280 275 273 273 264 Mean fixation durations (ms) Mean fixation durations (ms) 260 260 248 248 243 243 240 240 234 230 228 224 220 220 200 200 ST TT Dictionary ST TT Dictionary PA Text 1 Text 2 Text 1 Text 2 Mean fixation durations by Text Type in different AOIs (seconds). Mean fixation durations by Text Type in different subtasks (seconds).

1800 1800 Mean pupil size (arbitrary units) Mean pupil size (arbitrary units) 1615 1601 1615 1591 1600 1600 1518 1503 1501 1518 1497 1487 1503 1488 1389 1389 1400 1400 1200 1200 1000 1000 800 800 ST TT Dictionary PA ST TT Dictionary Text 1 Text 2 Text 1 Text 2 Mean pupil size by Text Type in different AOIs (arbitrary units). Mean pupil size by Text Type in different AOIs (arbitrary units).

278 280 1800 275 273 Mean pupil size (arbitrary units) 1615 1601 Mean fixation durations (ms) 264 1600 1518 1503 1501 1497 260 1487 248 1389 243 1400 240 230 1200 224 220 1000 200 800 ST TT Dictionary PA ST TT Dictionary PA Text 1 Text 2 Text 1 Text 2 Mean pupil size by Text Type in different AOIs (arbitrary units). Mean fixation durations by Text Type in different subtasks (seconds).

400 1600 1529 340 Mean pupil size (arbitrary units) 329 350 Mean fixation durations (ms) 1404 1395 300 1400 1339 237 234 250 200 1200 150 100 1000 50 0 800 Dictionary without key events Dictionary with key events Text 1 Text 2 Text 1 Text 2 Dictionary without key events Dictionary with key events Mean durations of fixations in dictionary area with/ without key events Mean pupil size in dictionary area with/ without key events being detected (seconds). being detected (arbitrary units).

Cross AOIs saccades ST-TT ST-Dict TT-Dict Total Text Type Text 1 Count 1210 158 232 1600 Percentage 75.6% 9.9% 14.5% 100.0% Text 2 Count 1044 170 218 1432 Percentage 72.9% 11.9% 15.2% 100.0% Total Count 2254 328 450 3032 Percentage 74.3% 10.8% 14.8% 100.0%

 Assuming different reading patterns 265 can be identified due to distinctive 261 needs for the subtask 260 Mean fixation durations (ms) 255  Calculating duration of fixations after 252 251 250 cross-interest-area saccades 245  No significance founded 240 240 235  Limitations 230 ST-Dictionary TT=Dictionary  “Micro - cycle” of translation by Jakobsen Text 1 Text 2  No small-scale text analysis conducted in Mean duration of fixations in dictionary area after ST-Dictionary and this project TT-Dictionary saccades (ms).

 To investigate and map out a generalized pattern of translation process, and propose possible avenues for future researches  Mainly three aspects of translation process (general distribution of cognitive attentions, cognitive workload of various subtasks and working style of translators) were investigated

 Overall distribution (dwell time)  TT > ST > Dictionary  No main effect from text complexity  Cognitive workload (fixation duration; pupil size)  TT production, ST comprehension ?> Dictionary, PA  Pupil size has positive relationship with text complexity in ST, TT, Dict  More discreet identification of subtasks required  Working style  Shift frequency: ST-TT > ST/TT-Dictionary  Dictionary reading pattern: no significance