Hacettepe University Graduate School of Social Sciences Department of Translation and Interpreting Translation and Interpreting in English Programme A SOCIOLOGICAL PERSPECTIVE ON THE DIGITALISATION OF THE TRANSLATION PROCESS: THE EFFECTS ON THE OCCUPATIONAL STATUS OF TRANSLATORS Büşra KURT UÇAR Ph.D. Dissertation Ankara, 2025 A SOCIOLOGICAL PERSPECTIVE ON THE DIGITALISATION OF THE TRANSLATION PROCESS: THE EFFECTS ON THE OCCUPATIONAL STATUS OF TRANSLATORS Büşra KURT UÇAR Hacettepe University Graduate School of Social Sciences Department of Translation and Interpreting Translation and Interpreting in English Programme Ph.D. Dissertation Ankara, 2025 ACCEPTANCE AND APPROVAL The jury finds that Büşra KURT UÇAR has on the date of 20 June 2025 successfully passed the defense examination and approves her Doctoral Dissertation titled “A Sociological Perspective on the Digitalisation of the Translation Process: The Effects on the Occupational Status of Translators”. Assoc. Prof. Dr. Hilal ERKAZANCI DURMUŞ (Jury President) Assoc. Prof. Dr. Sinem BOZKURT (Main Adviser) Prof. Dr. Mehmet ŞAHİN Assist. Prof. Dr. Duygu DUMAN Assist. Prof. Dr. Alper KUMCU I agree that the signatures above belong to the faculty members listed. Prof. Dr. Uğur ÖMÜRGÖNÜLŞEN Graduate School Director YAYIMLAMA VE FİKRİ MÜLKİYET HAKLARI BEYANI Enstitü tarafından onaylanan lisansüstü tezimin tamamını veya herhangi bir kısmını, basılı (kağıt) ve elektronik formatta arşivleme ve aşağıda verilen koşullarla kullanıma açma iznini Hacettepe Üniversitesine verdiğimi bildiririm. Bu izinle Üniversiteye verilen kullanım hakları dışındaki tüm fikri mülkiyet haklarım bende kalacak, tezimin tamamının ya da bir bölümünün gelecekteki çalışmalarda (makale, kitap, lisans ve patent vb.) kullanım hakları bana ait olacaktır. Tezin kendi orijinal çalışmam olduğunu, başkalarının haklarını ihlal etmediğimi ve tezimin tek yetkili sahibi olduğumu beyan ve taahhüt ederim. Tezimde yer alan telif hakkı bulunan ve sahiplerinden yazılı izin alınarak kullanılması zorunlu metinleri yazılı izin alınarak kullandığımı ve istenildiğinde suretlerini Üniversiteye teslim etmeyi taahhüt ederim. Yükseköğretim Kurulu tarafından yayınlanan “Lisansüstü Tezlerin Elektronik Ortamda Toplanması, Düzenlenmesi ve Erişime Açılmasına İlişkin Yönerge” kapsamında tezim aşağıda belirtilen koşullar haricince YÖK Ulusal Tez Merkezi / H.Ü. Kütüphaneleri Açık Erişim Sisteminde erişime açılır. o Enstitü / Fakülte yönetim kurulu kararı ile tezimin erişime açılması mezuniyet tarihimden itibaren 2 yıl ertelenmiştir. (1) o Enstitü / Fakülte yönetim kurulunun gerekçeli kararı ile tezimin erişime açılması mezuniyet tarihimden itibaren ….. ay ertelenmiştir. (2) o Tezimle ilgili gizlilik kararı verilmiştir. (3) 17/07/2025 Büşra KURT UÇAR 1 “Lisansüstü Tezlerin Elektronik Ortamda Toplanması, Düzenlenmesi ve Erişime Açılmasına İlişkin Yönerge” (1) Madde 6. 1. Lisansüstü tezle ilgili patent başvurusu yapılması veya patent alma sürecinin devam etmesi durumunda, tez danışmanının önerisi ve enstitü anabilim dalının uygun görüşü üzerine enstitü veya fakülte yönetim kurulu iki yıl süre ile tezin erişime açılmasının ertelenmesine karar verebilir. (2) Madde 6. 2. Yeni teknik, materyal ve metotların kullanıldığı, henüz makaleye dönüşmemiş veya patent gibi yöntemlerle korunmamış ve internetten paylaşılması durumunda 3. şahıslara veya kurumlara haksız kazanç imkanı oluşturabilecek bilgi ve bulguları içeren tezler hakkında tez danışmanının önerisi ve enstitü anabilim dalının uygun görüşü üzerine enstitü veya fakülte yönetim kurulunun gerekçeli kararı ile altı ayı aşmamak üzere tezin erişime açılması engellenebilir. (3) Madde 7. 1. Ulusal çıkarları veya güvenliği ilgilendiren, emniyet, istihbarat, savunma ve güvenlik, sağlık vb. konulara ilişkin lisansüstü tezlerle ilgili gizlilik kararı, tezin yapıldığı kurum tarafından verilir *. Kurum ve kuruluşlarla yapılan işbirliği protokolü çerçevesinde hazırlanan lisansüstü tezlere ilişkin gizlilik kararı ise, ilgili kurum ve kuruluşun önerisi ile enstitü veya fakültenin uygun görüşü üzerine üniversite yönetim kurulu tarafından verilir. Gizlilik kararı verilen tezler Yükseköğretim Kuruluna bildirilir. Madde 7.2. Gizlilik kararı verilen tezler gizlilik süresince enstitü veya fakülte tarafından gizlilik kuralları çerçevesinde muhafaza edilir, gizlilik kararının kaldırılması halinde Tez Otomasyon Sistemine yüklenir. * Tez danışmanının önerisi ve enstitü anabilim dalının uygun görüşü üzerine enstitü veya fakülte yönetim kurulu tarafından karar verilir. ETİK BEYAN Bu çalışmadaki bütün bilgi ve belgeleri akademik kurallar çerçevesinde elde ettiğimi, görsel, işitsel ve yazılı tüm bilgi ve sonuçları bilimsel ahlak kurallarına uygun olarak sunduğumu, kullandığım verilerde herhangi bir tahrifat yapmadığımı, yararlandığım kaynaklara bilimsel normlara uygun olarak atıfta bulunduğumu, tezimin kaynak gösterilen durumlar dışında özgün olduğunu, Doç. Dr. Sinem BOZKURT danışmanlığında tarafımdan üretildiğini ve Hacettepe Üniversitesi Sosyal Bilimler Enstitüsü Tez Yazım Yönergesine göre yazıldığını beyan ederim. Büşra KURT UÇAR iv ACKNOWLEDGEMENTS Completing this Ph.D. journey has been one of the most challenging and rewarding experiences of my life. This dissertation would not have been possible without the guidance, support and encouragement of many people to whom I owe a debt of gratitude. First and foremost, I am sincerely thankful to my advisor, Assoc. Prof. Dr. Sinem Bozkurt, for her invaluable guidance and support throughout every stage of this process. Her thoughtful mentorship and dedication to academic rigor have contributed my development as a researcher in profound ways. I am especially grateful to her for always challenging me to think critically and aim higher. I would also like to express my sincere appreciation to the members of my thesis committee, Assoc. Prof. Dr. Hilal Erkazancı Durmuş and Prof. Dr. Mehmet Şahin, for their insightful feedback and generous engagement with my thesis. I feel fortunate to have had their guidance. I am further grateful to the jury members, Asst. Prof. Dr. Duygu Duman and Asst. Prof. Dr. Alper Kumcu, for taking the time to evaluate my thesis in detail and for offering their precious perspectives. Their expertise has significantly enriched this dissertation. A special thank you goes to my dear colleague and academic companion, Zeynep Şengel. Her fellowship, support and shared sense of purpose have been a constant source of motivation. From long discussions and moments of doubt to shared achievements, her presence made this journey more manageable. I would also like to thank my translator colleagues for their encouragement, understanding and tolerance throughout this long journey. I would like to extend my deepest gratitude to my family, whose love and belief in me have sustained me throughout this process. To my mother and father, to my sister and brother, for their constant encouragement; thank you for always being there for me. I would also like to thank my mother-in-law for taking care of me; thank you for your support. v Above all, I am profoundly grateful to my husband, Engin Uçar, whose unwavering support, patience, and love have been my anchor. He stood by me through every challenge, celebrated every small victory and carried me through moments of doubt and exhaustion. This journey would not have been possible without his love and support. To all of you, thank you. vi ABSTRACT KURT UÇAR, Büşra. A Sociological Perspective on the Digitalisation of the Translation Process: The Effects on the Occupational Status of Translators, Ph.D. Dissertation, Ankara, 2025. Digitalisation has changed and continues changing the world in various aspects. The translation sector, being no exception, has been highly affected by the introduction, development and dissemination of translation-related digital technologies. It is argued that this evolving digital environment has radically and irreversibly affected many parameters which are generally accepted as factors that affect the occupational status. Digitalisation is claimed to be the arch f(actor) that impacts the status perceptions in relation to translation as a profession. To this end, this thesis aims to investigate the interplay between the social and the technical actors operating within the co-constructed sociotechnical network based on the actor-network theory (ANT) proposed by Callon, Latour and Law. How translators position themselves in their occupational networks, how actors translate each other and how technical actors participate in the social order were scrutinised conducting questionnaires and interviews with freelance translators, institutional translators and the respective employers, project managers or translation bureau owners in Turkey. It is revealed that digital translation technologies are involved in every translator’s network and digitalisation is a critical (f)actor that affects all status parameters to a certain extent. Furthermore, it was concluded that as long as translators could position themselves and their digital skills as the obligatory passage points, the network could maintain its stability, contributing to higher status perceptions. Developing an understanding of the occupational status that is shaped by the effects of digitalisation may raise awareness among translators and other actors, lead the prospective and practising translators to reframe their occupational identity and acquire new skill sets to maintain the stability of their occupational networks and help the translation and interpreting departments tailor their structures to adapt to the changing market and attract future translators. Keywords digitalisation, occupational status of translators, actor-network theory, sociology of translators, sociology of translation, obligatory passage point, digital skills vii ÖZET KURT UÇAR, Büşra. Sosyolojik Bir Bakış Açısıyla Çeviri Sürecinde Dijitalleşme: Dijitalleşmenin Çevirmenlerin Mesleki Statüsüne Etkileri, Doktora Tezi, Ankara, 2025. Dijitalleşme tüm dünyada çeşitli değişikliklere sebep olmuştur ve olmaya da devam etmektedir. İstisnai olmayan çeviri sektörü ise çeviri ile ilgili dijital teknolojilerin piyasaya tanıtılması, geliştirilmesi ve yaygınlaşmasından önemli düzeyde etkilenmiştir. Değişen dijital ortamın, mesleki statüyü etkilediği kabul edilen pek çok parametreyi kökten ve geri döndürülemez bir şekilde etkilediği savunulmaktadır. Dijitalleşmenin bir meslek olarak çevirmenlik ile ilgili statü algılarını etkileyen ana (f)aktör olduğu ileri sürülmektedir. Bu doğrultuda, bu tezde ortaklaşa inşa edilmiş sosyoteknik ağda faaliyet gösteren sosyal ve teknik aktörler arasındaki etkileşimin Callon, Latour ve Law tarafından ortaya konan aktör-ağ kuramı (AAK) ışığında araştırılması amaçlanmaktadır. Çevirmenlerin mesleki ağlarında kendilerini nasıl konumlandırdıkları, aktörlerin birbirini nasıl dönüştürdüğü ve teknik aktörlerin sosyal düzene nasıl katıldığı Türkiye’deki serbest çevirmenler, kurum çevirmenleri ve ilgili işverenler, proje yöneticileri veya çeviri bürosu sahipleri ile yapılan anketler ve görüşmeler aracılığıyla ayrıntılı şekilde incelenmiştir. Dijital çeviri teknolojilerinin tüm çevirmenlerin ağına dâhil olduğu ve dijitalleşmenin tüm statü parametrelerini belirli bir noktaya kadar etkileyen kritik bir (f)aktör olduğu ortaya konmuştur. Ayrıca çevirmenler kendilerini ve dijital becerilerini zorunlu geçiş noktası olarak konumlandırabildiği sürece ağın stabilitesini koruyabileceği, bunun da daha yüksek bir statü algısına katkıda bulunabileceği sonucuna varılmıştır. Mesleki statünün dijitalleşmenin etkisiyle şekillenen bir kavram olarak anlaşılması; çevirmenler ve diğer aktörlerin farkındalığının artmasına katkıda bulunabilir, potansiyel ve bilfiil çalışan çevirmenlerin mesleki ağlarının stabilitesini sürdürmek için mesleki kimliklerini yeni bir çerçeveye oturtup yeni beceri setleri edinmelerine destek olabilir ve mütercim ve tercümanlık bölümlerinin değişen piyasaya uyum sağlamasına ve geleceğin çevirmenlerini bu alana yönlendirebilmesine yardımcı olabilir. Anahtar Sözcükler dijitalleşme, çevirmenlerin mesleki statüsü, aktör-ağ kuramı, çevirmen sosyolojisi, dönüştürüm sosyolojisi, zorunlu geçiş noktası, dijital beceriler viii TABLE OF CONTENTS ACCEPTANCE AND APPROVAL .............................................................................. i YAYIMLAMA VE FİKRİ MÜLKİYET HAKLARI BEYANI ................................. ii ETİK BEYAN ................................................................................................................ iii ACKNOWLEDGEMENTS .......................................................................................... iv ABSTRACT ................................................................................................................... vi ÖZET ............................................................................................................................. vii TABLE OF CONTENTS ............................................................................................ viii ABBREVIATIONS ..................................................................................................... xiii LIST OF TABLES ........................................................................................................ xv LIST OF FIGURES .................................................................................................... xvi INTRODUCTION ........................................................................................................... 1 CHAPTER 1: DIGITALISATION .............................................................................. 11 1.1. DIGITALISATION WITHIN THE CONTEXT OF TRANSLATION STUDIES ................................................................................................................... 11 1.2. MACHINE TRANSLATION (MT) AND RELATED CONCEPTS ............ 13 1.2.1. Machine Translation (MT) ........................................................................... 13 1.2.1.1. Direct approach ..................................................................................... 18 1.2.1.2. Rule-based approaches ......................................................................... 19 1.2.1.2.1. Interlingua approach ...................................................................... 19 1.2.1.2.2. Transfer approach .......................................................................... 20 1.2.1.3. Corpus-based approaches ..................................................................... 20 1.2.1.3.1. Statistical-based approach ............................................................. 20 1.2.1.3.2. Example-based approach ............................................................... 21 1.2.1.3.3. Neural-based approach .................................................................. 22 1.2.2. Online MT systems ...................................................................................... 23 1.3. COMPUTER-AIDED TRANSLATION (CAT) AND RELATED CONCEPTS............................................................................................................... 26 1.3.1. Translation Memories and Termbases ......................................................... 28 1.3.2. Online CAT Systems .................................................................................... 31 1.4. HISTORICAL BACKGROUND OF THE TRANSLATION-RELATED DIGITAL TOOLS .................................................................................................... 32 ix 1.4.1. The History of Machine Translation (MT) Systems .................................... 33 1.4.2. The History of Computer-Aided Translation (CAT) Tools ......................... 41 1.5. THE EFFECTS OF THE DIGITALISATION ............................................... 49 1.5.1. Practice-based effects of the digitalisation ................................................... 50 1.5.1.1. Effects on microcosm of translators ..................................................... 50 1.5.1.2. Effects on macrocosm of translators ..................................................... 54 1.5.2. Theory-based effects of the digitalisation .................................................... 56 1.5.3. Technology-enabled and technology-focused research ............................... 60 1.6. THE TECHNOLOGICAL TURN ................................................................... 62 1.7. THE SITUATION IN TURKEY IN RELATION TO TRANSLATION- RELATED DIGITALISATION .............................................................................. 68 CHAPTER 2: THE SOCIOLOGY OF TRANSLATION ......................................... 75 2.1. STRUCTURE/AGENCY DICHOTOMY ....................................................... 75 2.2. THE SOCIOLOGICAL TURN IN TRANSLATION STUDIES .................. 77 2.2.1. Sociological Studies in the Field .................................................................. 79 2.2.2. The Intersection of Digitalisation and Human Issues .................................. 82 2.2.2.1. Dance of Agency .................................................................................. 85 2.2.3. Translation as a Social Practice .................................................................... 87 2.3. THE SOCIAL CONSTRUCTION OF TECHNOLOGY (SCOT) ................ 89 2.4. ACTOR-NETWORK THEORY (ANT) .......................................................... 91 2.4.1. Actor and Network ....................................................................................... 93 2.4.2. Concept and Process of Translation in ANT ................................................ 98 2.4.3. ANT in Translation Studies ....................................................................... 102 2.4.3.1. ANT in a Turkish Context .................................................................. 105 CHAPTER 3: THE OCCUPATIONAL STATUS OF TRANSLATORS ............. 106 3.1. PROFESSION OR OCCUPATION? ............................................................ 106 3.1.1. The Intersection of the Digitalisation and the Future of the Profession..... 110 3.1.1.1. Skilled-Biased Technological Change ................................................ 111 3.2. THE CONCEPTS OF STATUS AND OCCUPATIONAL STATUS ......... 112 3.2.1. Why Occupational Status Should Be Studied? .......................................... 114 3.2.2. The Professional/Occupational Status of Translators/Interpreters ............. 116 3.3. THE OCCUPATIONAL STATUS OF TRANSLATORS IN TURKEY ... 122 x CHAPTER 4: METHODOLOGY ............................................................................. 127 4.1. CASE STUDY DESIGN .................................................................................. 127 4.2. PARTICIPANTS ............................................................................................. 127 4.3. DATA COLLECTION .................................................................................... 129 4.4. DATA ANALYSIS ........................................................................................... 133 4.4.1. Questionnaires 1 and 2 ............................................................................... 133 4.4.2. Interviews ................................................................................................... 134 4.5. ETHICAL CONSIDERATIONS AND LIMITATIONS ............................. 135 CHAPTER 5: ANALYSES AND FINDINGS .......................................................... 139 5.1. ANALYSIS OF THE QUESTIONNAIRES .................................................. 139 5.1.1. Analysis of Answers to Questionnaire 1 (For Translators) ........................ 140 5.1.1.1. Background Information and More Regarding Translators ................ 140 5.1.1.2. Translators’ Perceptions on Salary/Remuneration ............................. 150 5.1.1.3. Translators’ Perceptions on Education/Expertise ............................... 152 5.1.1.4. Translators’ Perceptions on Competitive Advantage/Quality Work/Continuous Development ...................................................................... 154 5.1.1.5. Translators’ Perceptions on Professional Exclusivity ........................ 158 5.1.1.6. Translators’ Perceptions on Power/Influence/Authority .................... 160 5.1.1.7. Translators’ Perceptions on Solidarity ................................................ 162 5.1.1.8. Translators’ Perceptions on Job Satisfaction ...................................... 164 5.1.1.9. Translators’ Perceptions on Recognition/Prestige .............................. 166 5.1.1.10. Translators’ Perceptions on Appreciation ........................................ 167 5.1.1.11. Translators’ Perceptions on the Future/Changing Roles/Ethics ....... 168 5.1.2. Analysis of Answers to Questionnaire 2 (For Employers) ........................ 185 5.1.2.1. Background Information and More Regarding Employers ................ 186 5.1.2.2. Employers’ Perceptions on Fees ......................................................... 189 5.1.2.3. Employers’ Perceptions on Education/Expertise ................................ 191 5.1.2.4. Employers’ Perceptions on Competitive Advantage/Quality Work/Continuous Development ...................................................................... 193 5.1.2.5. Employers’ Perceptions on Professional Exclusivity ......................... 197 5.1.2.6. Employers’ Perceptions on Power/Influence/Authority ..................... 198 5.1.2.7. Employers’ Perceptions on Appreciation ........................................... 200 xi 5.1.2.8. Employers’ Perceptions on the Future/Changing Roles/Ethics .......... 201 5.2. ANALYSIS OF QUALITATIVE DATA ....................................................... 208 5.2.1. Introducing Oneself Professionally ............................................................ 214 5.2.2. Definition of a Translator ........................................................................... 215 5.2.3. Translation or Profession? And Why? ....................................................... 216 5.2.4. The Actors in the Occupational Networks and Their Roles ....................... 217 5.2.5. Use of Digital Translation Tools ................................................................ 222 5.2.6. How Digital Tools Affect the Future of the Translation Profession .......... 226 5.2.7. Prestige of the Translation Profession ........................................................ 229 5.2.8. Would You Recommend Someone To Be A Translator? .......................... 232 CHAPTER 6: DISCUSSION ..................................................................................... 234 6.1. ISSUE OF IDENTITY .................................................................................... 235 6.2. THE (F)ACTORS IN THE OCCUPATIONAL NETWORKS OF TRANSLATORS .................................................................................................... 237 6.2.1. Economics .................................................................................................. 239 6.2.1.1. Analyse Files Reports as Mediators ................................................... 243 6.2.1.2. A Strict Rule ....................................................................................... 245 6.2.2. Quality and Responsibility ......................................................................... 247 6.2.3. Monitoring and Transparency .................................................................... 249 6.3. PRESTIGIOUS OR NOT PRESTIGIOUS?: BEHIND THE SCENES ..... 250 6.3.1. Fear of Displacement ................................................................................. 253 6.3.2. The Future of the Profession ...................................................................... 254 6.3.2.1. A Unified Entity with Dissidents ........................................................ 256 CONCLUSION ............................................................................................................ 261 BIBLIOGRAPHY ....................................................................................................... 266 APPENDIX 1: CONSENT FORM ............................................................................ 283 APPENDIX 2: QUESTIONNAIRE 1 - QUESTIONNAIRE FOR TRANSLATORS ....................................................................................................................................... 284 APPENDIX 3: QUESTIONNAIRE 2 - QUESTIONNAIRE FOR EMPLOYERS ....................................................................................................................................... 298 APPENDIX 4: INTERVIEW QUESTIONS ............................................................. 307 APPENDIX 5: DEMOGRAPHICS OF PARTICIPANT TRANSLATORS ......... 310 xii APPENDIX 6: BACKGROUND INFORMATION, PRESTIGE PERCEPTION AND DIGITAL TOOL USE OF INTERVIEWEES ................................................ 314 APPENDIX 7: ORIGINALITY REPORT ............................................................... 317 APPENDIX 8: ETHICS COMMISSION FORM .................................................... 319 xiii ABBREVIATIONS AI: Artificial intelligence ANT: Actor-network theory CAT: Computer-aided tool EBMT: Example-based machine translation FT-ASI: Freelance translator whose additional source of income is translation FT-MSI: Freelance translator whose main source of income is translation GT: Google translate GNMT: Google neural machine translation HAMT: Human-aided machine translation IT: Institutional translator MAHT: Machine-aided human translation MAT: Machine-aided translation MT: Machine translation MTPE: Machine translation post-editing NMT: Neural machine translation RBMT: Rule-based machine translation RNN: Recurrent neural networks SCOT: Social construction of technology SL: Source language SMT: Statistical machine translation T&I: Translation and Interpreting xiv TB: Termbase TEnTs: Translation environment tools TL: Target language TM: Translation memory xv LIST OF TABLES Table 1: Comparison of ratings of ITs and FTs ........................................................... 148 Table 2: Comparison Based on Prestige Perceptions ................................................... 149 Table 3: Comments of Participant Translators ............................................................. 179 Table 4: Comparison of Ratings of Employers and Translators for the Same Statements ....................................................................................................................................... 185 Table 5: Comments of Participant Employers ............................................................. 207 xvi LIST OF FIGURES Figure 1: MT Architectures ............................................................................................ 18 Figure 2: Holmes’ map of Translation Studies .............................................................. 58 Figure 3: Quah’s extended version of the Applied branch ............................................. 59 Figure 4: Munday et al.’s extended version of the Applied branch ............................... 59 Figure 5: Age Range of Translators ............................................................................. 140 Figure 6: How Participant Translators Define Their Profession .................................. 141 Figure 7: Details of Profession Definitions .................................................................. 142 Figure 8: Years of Experience...................................................................................... 142 Figure 9: Number of Hours Allocated to Translation and Related Jobs ...................... 143 Figure 10: Monthly Income ......................................................................................... 144 Figure 11: Prestige of Translation Profession .............................................................. 145 Figure 11a: Comparison of Groups for Prestige Perceptions……………...……………...…146 Figure 12: Use of Digital Translation Tools ................................................................ 146 Figure 13: Date of Start to Use CAT Tools ................................................................. 147 Figure 14: Date of Start to Use MT/AI Tools .............................................................. 148 Figure 15: Advantage For More Income ...................................................................... 151 Figure 16: Negative Effects on Justification of Pricing ............................................... 151 Figure 17: Improbability of Replacement by MT/AI software .................................... 152 Figure 18: Negative Effects on Completed Formal Education .................................... 153 Figure 19: Inevitable Need to Use Digital Translation Tools ...................................... 154 Figure 20: Effects on Increasing the Volume of Work ................................................ 155 Figure 21: Tighter Deadlines ....................................................................................... 156 Figure 22: Preferability as an Employee ...................................................................... 157 Figure 23: Requirement to Keep the Mistakes Due to Locked Segments/References 158 Figure 24: “Anyone Can Do” Approach ...................................................................... 158 Figure 25: Showing the Difficulty of Translation ........................................................ 159 Figure 26: Negative Effects on the Value of Knowledge ............................................ 160 Figure 27: Decreased Decision-Making Authority ...................................................... 161 Figure 28: Feeling Controlled ...................................................................................... 161 Figure 29: Feeling Less Responsible for Post-Edited Texts ........................................ 162 Figure 30: Creating A Sense of Solidarity Among Translators ................................... 163 Figure 31: Usefulness in Disaster Situations ............................................................... 164 Figure 32: Positive Effects on Job Satisfaction............................................................ 165 Figure 33: Desire to Work in a Digital Environment ................................................... 165 Figure 34: Decreased Respect for Expertise ................................................................ 166 Figure 35: Increasing the Importance of Translators ................................................... 167 Figure 36: Reducing the Level of Appreciation ........................................................... 168 Figure 37: Vulnerability Against AI ............................................................................ 169 Figure 38: Change of Roles into Post-Editors.............................................................. 170 Figure 39: Negative Effects on Translator Identity...................................................... 170 xvii Figure 40: Reduced Need for Translators due to AI .................................................... 171 Figure 41: Positive Effects on the Reputation of Translation Profession .................... 172 Figure 42: Ending the Profession ................................................................................. 172 Figure 43: Reduced Value of Profession due to Crowdsourcing ................................. 173 Figure 44: Feeling of Violating the Rights of Other Translators ................................. 174 Figure 45: Violation of Ownership Rights due to Online Platforms ........................... 175 Figure 46: Feeling Under Surveillance on Online Platforms ....................................... 176 Figure 47: Satisfaction with Online Platforms ............................................................. 177 Figure 48: Age Range of Employers/Project Managers/Translation Bureau Owners . 186 Figure 49: Participants’ Definition of Their Profession ............................................... 187 Figure 50: Prestige of Translation Profession .............................................................. 188 Figure 51: Requesting Translators to Use CAT tools .................................................. 188 Figure 52: Using MT/AI Tools .................................................................................... 189 Figure 53: Increase in Translation/Proofreading Fees ................................................. 190 Figure 54: Increase in Post-editing Fees ...................................................................... 190 Figure 55: Improbability of Total Replacement by MT/AI ......................................... 191 Figure 56: Improbability of Delivering MT/AI Translations without Post-editing ..... 192 Figure 57: Trivialising the Translators’ Knowledge .................................................... 192 Figure 58: Inevitable Need to Use Digital Tools in Translation Jobs .......................... 193 Figure 59: Not Preferring to Work With Translators Who Do Not Use CAT Tools ... 194 Figure 60: MT/AI’s Facilitating Translation................................................................ 194 Figure 61: MT/AI’s Accelerating Translation ............................................................. 195 Figure 62: Unimportance of MT/AI Use If the Final Product is Quality..................... 196 Figure 63: Insufficiency for Client Expectations ......................................................... 196 Figure 64: “Anyone Can Do” Approach ...................................................................... 197 Figure 65: Showing the Difficulty of Translation ........................................................ 198 Figure 66: Monitoring Translators via Online Platforms ............................................. 199 Figure 67: Keeping the Post-Editor Responsible ......................................................... 199 Figure 68: Praise to the Developers ............................................................................. 200 Figure 69: Role of Translators in Developing Tools ................................................... 201 Figure 70: Decreased Need for Translators ................................................................. 201 Figure 71: Need for Translators as Post-Editors .......................................................... 202 Figure 72: Increased Importance of the Profession...................................................... 203 Figure 73: Making the Profession Less Valuable ........................................................ 203 Figure 74: Ending the Profession ................................................................................. 204 Figure 75: Non-necessity of Informing the Client About MT/AI Use......................... 205 Figure 76: Non-necessity of Being Informed About the Use of MT/AI ...................... 206 Figure 77: Categories Obtained After Thematic Analysis in MAXQDA 24 ............... 209 Figure 78: The Frequencies of Main Categories in the Code Matrix Browser ............ 210 Figure 79: Breakdown of Sub-Categories Analysed in MAXQDA - 1 ....................... 212 Figure 80: Breakdown of Sub-Categories Analysed in MAXQDA - 2 ....................... 213 Figure 81: Who Translator Is ....................................................................................... 215 xviii Figure 82: Reasons Behind Definitions as Profession and Occupation ....................... 216 Figure 83: Actors in Occupational Networks of Interviewees ..................................... 218 Figure 84: Declaring the Use of MT/AI ....................................................................... 226 Figure 85: Optimism and Pessimism at Present and for the Future ............................. 227 Figure 86: If Digital Translation Tools Disappeared From the Market ....................... 231 Figure 87: Recommending Being a Translator ............................................................ 232 Figure 88: The Most Common Actors in the Occupational Networks of Freelancers . 238 1 INTRODUCTION Technology advances at an unprecedented pace and the effects of digitalisation are increasingly observed on many aspects of life and human activities. Most business processes are irreversibly affected while ways of doing business, methods, connections, interactions between parties and dependence on the human labour constantly change. The translation services industry is no exception with its automated or semi-automated processes. Nonetheless, digital technologies encompass numerous tools, software and utilities from computers to operating systems, from text editors to word processors, from online dictionaries to machine translation (MT) and artificial intelligence (AI) and computer-aided translation (CAT) tools. However, some of these technological innovations are directly for translation purposes while the others are not despite the fact that they are often used and mostly necessary for translation tasks. However, non- translation-directed technologies such as grammar checkers, word processors, spell checkers or search tools and technologies for interpreters such as MT systems for speech recognition, often in a restricted domain, will not be specifically focused in this thesis but they will be surely touched upon as necessary for a complete description, which means the translated-related digital tools that would be investigated refer to the technological developments after mid-1990s. Development and employment of the tools, programs and devices for translation purposes drastically changed the face of translation sector all around the world. In the market, it is mostly not an option but a requirement to use digital tools while performing translation or translation-related tasks at present. All of the changes driven by digitalisation point out major implications for translation practices which eventually lead to the born of a “technological turn” (Cronin, 2010, p. 1) in Translation Studies. Cronin attributes the emergence of this turn not to theoretical developments but to the changing translation practices and draws attention to the need to re-examine the position of translators (ibid.), which demonstrates a strong connection between technological developments and social factors that surround translators in their professional life. Lagoudaki (2008), in her study where she investigates the MT function as integrated in TM tools, states that the translator’s role may be reduced to that of a post-editor which may eventually lead to changes of translators’ roles. In his 2 study, Çetiner (2021) puts forward that the professional translation workflow has changed as a result of the improvements in machine translation technologies and suggests that the competences and skills of translators should be redefined, underlying the changing role of translators. As Şahin and Gürses (2021, p. 197) note in their study regarding literary translation through machine translation tools and post-editing processes, interaction between human and machine may have caused participants to feel some degree of confusion about their roles. With this confusion in mind, whether the ones who cannot keep up with continuously evolving digital tools and devices may be or will be able to survive in the translation market seems uncertain. In addition, even the ones who (can) follow the innovations may be still at risk of facing a changed definition of their roles and positions, which may lead them to question the profession of translation, the existence of the profession, their own capabilities, and possibly their career choices. Since digital translation tools are to be used by human translators, it is almost inevitable that the studies focusing on technology have implications on translators and/or interpreters as well. Accordingly, it may be claimed that this changing environment has a huge effect on translator’s occupational status. To this end, it is important to define the concept of status and occupational status as addressed in this study. Status refers to “social ranking” or “social prestige” in line with Wadensjö’s (2011) and Treiman’s (2004) arguments; accordingly the concept of occupational status is addressed as “social ranking based on one’s occupation” or “occupational prestige”; however, not just as perceived by the public (respective employers in this case) but also as perceived by the translators themselves (as measured by Dam and Zethsen, 2008). However, it should be noted that “[…] status cannot be viewed as an absolute notion but is a complex, subjective and context-dependent construct” (Dam and Zethsen, 2008, p. 74). Dam and Zethsen also state that categories including, but not limited to, income, educational background, class, race, gender, worthiness of the relevant role and prestige, determine which professions hold a high status and which ones hold a low status, noting that these parameters may change from one country to another (ibid.) and this encourages further studies with different participants (and also different characteristics) and in different settings. Moreover, the perceived professional status of translators affects their working conditions (ibid., p. 94), meaning that the situation is not one-sided but reciprocal. This shows that how translators see themselves has an 3 effect on their own working conditions and also the more challenging the conditions, the lower the status as perceived by translators themselves, creating a form of vicious circle. Dam and Zethsen (2008, p. 94) also underlines a very significant point, stating that status influences the very future of the profession as prospective translators may only be attracted to a translation-related career if the profession has a high status among both translators/interpreters and also in the society. At this point, the professional status of translators is already considered low, or middling at best, and the profession is accordingly regarded as low paid, insignificant, unappreciated or isolated (see Dam and Zethsen, 2008 and Ruokonen, 2016). Studying status in its general sociological sense is also worthwhile as our beliefs and perceptions have a strong impact on the way we act in the society and think, and thus, studying translator status would provide insights into the role of translators (Ruokonen, 2016, p. 188). Where the translators position themselves with respect to their profession in the social world may only be identified by investigating their working conditions, occupational identities, roles and interactions with other actors involved in the workflow and daily translation-related practices. It is obvious that digital technologies have their own benefits and advantages, most of which are probably indispensable for many translators and employers in the market; however, if translators are all contented, or aware of the threats, or concerned for their future has not been addressed in light of actor-network theory (ANT) focusing on the relation between digitalisation and the concept of occupational status in the literature yet. Moreover, the researchers in Translation Studies have not much leveraged from the ANT in their studies. This thesis will emphasise the value of this sociological theory and will demonstrate how it may be employed in and provide significant insights for Translation Studies not only on paper but also in the real world. In addition, it is possible that while translators may be aware of concrete effects of the digitalisation such as changes in their income, they may be unaware of their altering roles and other changes regarding more abstract parameters such as appreciation, recognition or job satisfaction. Thus, the results of this thesis may raise awareness of translators with regard to more subtle and abstract and yet even more important conditions to maintain their profession. Besides, understanding the effects of the 4 digitalisation on translator status will also guide the translation and interpreting (T&I) departments for revisiting or reorganising their structures. This thesis will shed light on the striking effects of the digital technologies in relation to translation on the daily professional life of translators within the sample set in Turkey. How the changes in roles, visibility, power, occupational identity, expertise or income that are caused primarily by digitalisation eventually and radically affect the status perception of translators as evaluated by translators themselves and employers will be demonstrated in order to raise awareness, especially for the future of the profession. In this similar vein, Fantinuoli (2019, p. 13-14) urgently calls for research and empirical evidence to foresee future trends in the field, which will prepare the next generations for the disruptive effects of technology. In addition to these, Dam and Zethsen underline that while researchers have concentrated very little on the translator in Translation Studies, there is even hardly ever any research focusing on translator status “as a subject in its own right” (2008, p. 72-73). Furthermore, translation is referenced as a “low- status profession” in the literature; however, the statements are not very solid and justified and there is a lack of empirical studies that address the topic extensively (Dam and Zethsen, 2008, p. 71-73). Since 2008, some studies were conducted on status but as Ruokonen (2016, p. 188) highlights the low-status of translation profession “was long taken for granted rather than considered a research topic”. Likewise, there are only a few studies (even fewer within the context of Turkey) discussing the topic of status, which is actually a very important sociological notion. Additionally, occupational status of translators has not been addressed in relation with both social and technical actors involved, to the best of my knowledge. Considering this gap, in this thesis, interactions of human and non-human actors will be addressed in light of the actor-network theory (ANT). These actors will be traced via questionnaires and interviews conducted on freelance translators, institutional translators and the respective employers, project managers or translation bureau owners in Turkey. This study aims at compensating for the potential disadvantages of questionnaires by in-depth interviews. All data collected in this study will be evaluated based on ANT as it highlights the interconnectedness and heterogeneity of the actors that may include the social as well as 5 the technical and may explain the potential interactions with its adaptable framework. It was first developed in 1980s by Bruno Latour, Michel Callon and John Law “to analyse situations in which it is difficult to separate humans and non-humans, and in which the actors have variable forms and competencies” (Callon, 2007, p. 274). ANT is considered appropriate and fit for the topic of this thesis particularly because digitalisation—common use of technological tools, programs, devices for performing translation and translation-related tasks—may be regarded as an actor based on this theory as ANT assigns agency to non-humans (i.e. all of the digital hardware/software professionally employed by translators) as well as humans (i.e. the translators and the employers). In ANT, the scholars call the central notion an actor or an actant, which includes both humans and non-humans such that humans interact with technical devices, machines or technological tools and all of these agents form the sociotechnical network in which they work and produce (Chesterman, 2006, p. 22). As underlined by Chesterman, translators complete their tasks through cooperation in the networks and each actor involved in these networks has a specific function or role, and more importantly, “each role has a status” (ibid., p. 23). Further explaining the relation between the technical and the social, Law and Callon state that: […] we are concerned to map the way in which [actors] define and distribute roles, and mobilize or invent others to play these roles. Such roles may be social, political, technical, or bureaucratic in character; the objects that are mobilized to fill them are also heterogeneous and may take the form of people, organizations, machines, or scientific findings. (1988, p. 285, italics in the original) From the point of view of ANT, each translation task is performed under certain constraints (Chesterman, 2006, p. 23) which may be, in this case, client requirements regarding the use of digital tools, obligation to accept and approve the translated segments or strings generated by non-human actors, inability to intervene with machine- translated and locked segments, the expectation of quality with even tighter deadlines. All such constraints pose challenges that may cause translators and employers question the translation profession, expertise, abilities, roles, identities and eventually the occupational status of translators. To explore these implications, it seems best to conduct an analysis based on the “sociology of translation” as suggested and used by Callon, describing the problematisation, interessement, enrolment, and mobilisation of 6 allies phases of this translation process in which “the identity of actors […] are negotiated and delimited” (1984, p. 203). Throughout this actor-tracing process, it is inevitable to encounter punctualised nodes (Callon, 1991 and Law, 1992), which may be explained as unified actors, the complex relations and network within which are not clearly visible or not needed to be known. These punctualised nodes, or black boxes (Latour, 2005) may be both human and non- human entities whose sociotechnical qualities are taken for granted and not questioned, but they may lose their integrity under certain circumstances. Tracing the actors, intermediaries and mediators and investigating Callon’s translation process1 will provide information and clues to reveal the effects of the digitalisation on the perceived occupational status of translators. At this point, digitalisation may be claimed to be one of the most influential actors, if not already at the top, in translation processes as performed and experienced by translators. According to ANT, as explained by Cressman: In any situation in which technology is used, it is used to delegate, or translate, a major effort into a minor effort. We delegate to technologies the work of many humans. In turn, technologies delegate behaviour back onto the social. We act as we do, not by some idealistic notion of free choice, but because our actions are bounded by technologies that delegate how and what we can do within a sociotechnical network. (2009, p. 10) This is particularly significant for underlining the limiting, constraining and steering effect of technologies within networks, which is completely in parallel with what the translators in the market live through every day. There is no reason to doubt that this evolving technological environment will have multifaceted effects on both translators and employers, which will, in return, cause an inevitable shift in the perceived occupational status of translators. Cronin acknowledges that the increased use of machine translation (MT) services would make the labour of translation invisible (2013, p. 99), which may most likely and eventually cause that the profession be in disgrace. It is argued that translators feel concerned with the rapid and irreversible overtake of 1 Translation process in Callon’s terms will be shown as translationANT process or translatedANT not to confuse it with translation process in translation studies, as inspired by Luo, W. (2020), throughout this thesis, where it is considered necessary. 7 digital technologies. This concern, however, is not mainly related to challenges of learning to use such technologies or keeping up with innovations but rather to the future, the questioned existence of the profession and the decreasing need for their services. Nonetheless, it should be noted that the power of this actor called digitalisation may not be felt at the same level by each actor. Therefore, its effects should be ideally compared between different parties with different conditions which will be described later. The research questions of this thesis are as follows: RQ1: Does the occupational network of human translators remain as a punctualised node until the interests or actions of actors that make it up diverge from the interests of translators? RQ2: Which actors are involved in translators’ occupational networks and how do these translate [according to Callon’s translation process] each other? RQ3: How do non-human translation-related digital technologies affect the occupational status of freelance translators and institutional translators in Turkey as perceived by themselves as well as employers? There are certain hypotheses of the researcher, which are as follows: H1: The occupational networks of translators are more stable when interests of all allies align and converge, which prevents punctualised nodes from disintegrating and also leads to perceptions of higher status. H2: Humans (translators or clients/employers) are mostly the OPPs that translate others and digital translation tools are mainly involved in translators’ networks as mediators. H3: Digitalisation in translation is the arch f(actor) that affects and shapes the occupational status of translators. In addition to the aforementioned arguments and justifications, providing a detailed structure of this thesis in detail will ensure ease of reference. In line with these explanations, Chapter 1 of this thesis first defines the concepts of digitalisation and 8 digitisation and explores digitalisation within the context of Translation Studies. It describes machine translation and related concepts, such as human-aided machine translation, pre-editing and post-editing, in detail while exploring the machine translation-related approaches developed to date. Afterwards, computer-aided translation and related concepts, such as translation memories and termbases, are described, explaining their functions and facilities. A sub-heading is devoted to the historical background of machine translation systems from the very first ideas to the recent developments, including important visionaries, companies and programs. This sub-heading is followed by another one that focuses on the history of computer-aided translation tools, highlighting the outstanding ideas argued so long ago as 1970s but had significant repercussions and were realised beyond expectations. This sub-heading shows how we, as translators, and the translation market have arrived this point in terms of digitalisation. Chapter 1 also investigates the practice-based effects of the digitalisation on the microcosm as well as macrocosm of translators. The concepts of Web 2.0, Web 3.0 (i.e. artificial intelligence technologies) and crowdsourcing are described in relation to their practical effects. The theory-based effects of the digitalisation are also addressed, detailing the technology-enabled and technology- focused research. The chapter continues with an in-depth description of technological turn in Translation Studies, justifying its existence and providing previous case studies in the literature to enlighten its effects. The chapter ends with the situation in Turkey considering digitalisation, exploring studies and theses in this field and listing technology-related courses in translation and interpreting departments in Turkey. Chapter 2 of this thesis begins with the sociological turn in Translation Studies, explaining the relation between sociology and Translation Studies. Effects and implications of this turn are explored through examples of studies that focus on sociology of translation from different perspectives applying Bourdieu’s, Luhmann’s or Latour, Callon and Law’s significant and originally social concepts and theories. This sub-heading is followed by the studies at the intersection of digitalisation, sociology and Translation Studies, highlighting the effects of the digitalisation on translators or interpreters. Also, concepts such as Pickering’s “mangle of practice” (1993) and “dance of agency” (1995) are described to further enrich the discussions. The concept of translation is explained as a social practice which leads to the heading Social 9 Construction of Technology (SCOT) to first explain the impact of society on technology before delving into the impact of technology on society. After these are elaborated in detail, actor-network theory (ANT), beginning from its birth, is described comprehensively, defining the key terms such as actor, actant, network, agency, translation, punctualisation, alignment, convergence, divergence, problematisation, interessement, enrolment, and mobilisation of allies. Finally, ANT in Translation Studies is addressed, shedding light on the need for further studies as there are essentially very few studies revolving around this intersection. Chapter 3 interrogates if translation is a profession or occupation. It underlines the relation between digitalisation and the future of the profession while describing the concept of skilled-biased technological change. The next heading explores the concept of status and occupational status, followed by the studies on professional/occupational status of translators/interpreters in the literature. These demonstrate the parameters that are used or revealed as affecting the perception of status. The chapter ends with the occupational status of translators in Turkey based on limited, but still illuminating, data. Chapter 4 explains the methodology of this study, beginning with the case study design. The profile of participants in the study are described in detail with inclusion and exclusion criteria. The methods and tools employed for data collection are elaborately described and the ethics board-approved questionnaire and interview questions were included at the end of the thesis as Appendices 2, 3 and 4. The details of determining the profile of participants and data collection are explained step by step. The chapter continues with the description of the tools that are used for data analysis, providing relevant reasons. Finally, ethical considerations with solutions to resolve possible problems, such as the ones regarding anonymity, are identified and limitations of the study are revealed, providing the possible reasons and justifications. Chapter 5 discusses the findings of research conducted within the scope of this thesis. Firstly, the responses of participants in Questionnaire 1 are evaluated collectively followed by the separate analysis of each sub-group, namely freelance translators and institutional translators. All of the information obtained through online questionnaires are discussed and the two sub-groups are compared based on the statistical data that include percentages and mean scores. The same evaluation is performed for the second 10 group that consists of employers, project managers and translation bureau owners in general based on their answers to Questionnaire 2. Narrative answers of the participants who accepted to be involved in semi-structured interviews are also analysed and evaluated. These three groups are compared with each other with an aim to reveal similarities and differences and the relevant reasons. Chapter 6 discusses all answers and statistical and qualitative data, considering their implications on the hypotheses and research questions of this study while also exposing the interactions between social and technical actors and tracing the relevant actor- networks within the scope of ANT. Conclusion chapter provides a structured but concise section on aims, purposes, findings, results and contributions of this study accompanied with suggestions for further research. 11 CHAPTER 1 DIGITALISATION In this chapter, the concept of digitalisation will be outlined and elaborated within the context of Translation Studies. Accordingly, the most known, employed and recognised translation-related digital tools and software will be described and exemplified, and their working principles, features and functions will be explained in detail. In addition, the definitions of often-confused terms, Machine Translation (MT), Machine-Aided Human Translation (MAHT), Human-Aided Machine Translation (HAMT), Computer- Aided Translation (CAT) and Machine-Aided Translation (MAT) will be made, and certain advantages and disadvantages of these systems will be highlighted. The specific approaches to MT will be elaborated in a chronological order from its birth, and the development of CAT systems will be elaborated with important dates. Afterwards, the effects of the digitalisation on the microcosm of translators will be described from different aspects ranging from the increased number of CAT and MT systems and their required use to issues of control over the translated product. The effects on the macrocosm of translators will also be elaborated based on the new ways of doing translation along with detailed explanations of Web 2.0 and Web 3.0 technologies and the concept of crowdsourcing. In addition, the indicators and effects of technological turn will be addressed both globally and in Turkey. The main reasons of such a detailed chapter is to conceptualise the subject matter, to clarify the often-confused terms and achieve a unanimity of terminology, and lastly to provide a comprehensive source of information as a contribution to the field. 1.1. DIGITALISATION WITHIN THE CONTEXT OF TRANSLATION STUDIES As the concept of digitalisation is often confused with digitisation, and these two concepts are mistakenly used interchangeably, it would be useful to define them before making an interpretation of the digitalisation within the context of Translation Studies. As Bloomberg (2018, p. 2) defines “digitization essentially refers to taking analogue information and encoding it into zeroes and ones so that computers can store, process, 12 and transmit such information” and it is a quite “straightforward term” that has a very clear meaning, a conversion. On the other hand, digitalisation does not have one single definition although the one by the Gartner, Inc. (an American IT company) has been recognised among many. Gartner defines digitalisation as “the use of digital technologies to change a business model and provide new revenue and value-producing opportunities; it is the process of moving to a digital business”2, which means it goes beyond and is more comprehensive than digitisation. Digitalisation leverages digital information and technologies to modify the business processes for efficiency and productivity, and it is “transforming the world of work” (Muro et al., 2017, p. 38) as well as changing work roles through automation (Bloomberg, 2018, p. 4). In addition, it does not only affect business models but lead to changes in society as well, because the development of the modern society heavily depends on digital technologies (Igolkin et al., 2020, p. 249-250). When it comes to Translation Studies, it may be claimed that digitalisation is heavily rooted in practice. However, digitalisation has not been theoretically conceptualised although a technological turn has been well accepted by the academic milieu. Considering the aforementioned definitions, it can be argued that translation processes have been modified and digitalised by digital technologies and the translation workflow is dramatically different from the way it used to be (pen and paper). Thanks to various technological advances over the decades, pen and paper were replaced with computers and machines and in the past 30 years they were replaced with sophisticated translation tools integrated with useful facilities. The translation world seems to have accepted that digital tools are now a requirement rather than an option in the market and the relationship between translation and technology is getting stronger each day, requiring the translator to adapt to the ever- changing environment (O’Hagan, 2013; Olohan, 2019; Marshman, 2014). In spite of this partly-forced popularity, there is still a general confusion around certain terms, namely machine translation (MT), human-aided/assisted machine translation (HAMT), machine-aided/assisted human translation (MAHT), computer-aided/assisted translation (CAT) and machine-aided/assisted translation (MAT). The differences among these terms mainly depend on the degree and extent of human- 2 See https://www.gartner.com/en/information-technology/glossary/digitalization https://www.gartner.com/en/information-technology/glossary/digitalization 13 machine/computer interaction involved during, before or after translation practice. These terms will be explained below as well as related concepts. 1.2. MACHINE TRANSLATION (MT) AND RELATED CONCEPTS Indeed, chronologically, it would be more appropriate to start with describing the concepts, types, former and recent approaches, and software related to machine translation (MT) as it dates back earlier than even the birth of computers. As Hutchins (2001, p. 5) claims, views of translators have polarised around two distinct extremes towards machine translation since the day the computers were regarded as practical facilities for translating languages in the 1940s and the first developments were obtained in research, particularly at various universities in the 1950s. The first group of translators were very comfortable because they never believed the idea that a machine could translate like a human translator while the second group was filled with trepidation foreseeing that they would lose their job to machines. It may be argued that the same controversial views are still applicable but probably not at these extremes rather with a more moderate approach towards cooperation and interaction between humans and machines. 1.2.1. Machine Translation (MT) The main term machine translation alone means that translation is only performed by machines (computerised systems) fully automatically without any human intervention. Essentially, a computer installed with some kind of a machine translation software or tool is required, and the human translator or non-translator has to only enter the source text in the designated space and give the computer the command of translation, which generates a raw translation. If the MT system is online, then no installed tool is actually necessary, and the user may go to the relevant link and write or copy and paste the source text in the designated window or just upload a file and has a target text within seconds. It should be noted that the human being does not involve in the actual translation process. When it comes to human-aided machine translation (HAMT), it is basically machine translation but with a slight difference. As Quah (2006, p. 173) elaborates, this is machine translation as the name indicates but allows human intervention during 14 translation in an interactive manner. Nonetheless, translation is still fundamentally performed by the machine itself. However, as the raw output does not mostly have the features of a high-quality translation performed by good translators, some additional steps were integrated into this machine translation process. There are two steps, one before feeding the machine with the source text -pre-editing- and the other is after the raw output is generated by the machine -post-editing. During pre-editing, the human being arranges the source text sentences in a way that is simple as much as possible, and easy to understand, and changes the vocabulary, when necessary (for instance homonyms may be replaced for clarity). Since the machine does not actually understand what a specific sentence or word means, the better the pre-editing step is managed with shorter, grammatically simple sentences the higher the chance of obtaining a quality output in the end. Post-editing, on the other hand, may require different skills from the ones necessary for pre-editing. Post-editing is quite similar to the processes of revision, editing or proofreading; however, in this case, the human being edits the translation generated by a machine not a human. While this post-editing may include the correction of simple typos, or misspellings, it may also necessitate the replacement of an entire sentence, translating from scratch, or searching for the accurate translation of key terms. Even though the logic behind the post-editing process is comparable to the editing process of a human translator’s work, Hutchins (1995, p. 431) notes that the mistakes made by MT systems are quite different than those of humans, such as pronoun, verb tense, preposition or article errors. A more recent study conducted by Daems et al. (2017) where they used Google Translate (in 2014) for translating English newspaper articles into Dutch showed that grammatical and syntactic errors such as word order and structural mistakes are the most common MT mistakes, followed by superfluous or omitted elements and wrong collocations. In addition to this, a more recent study carried out by Koçer Güldal and İşisağ (2019) for different text types from Turkish into English using SMT version of Google Translate, demonstrated that lexical errors were common in informative texts while semantic and pragmatic errors were included more heavily in operative and expressive texts. For neural MT, omission was the most common error (37%) from Chinese into English patent translation (Castilho et al., 2017). Nonetheless, these are not dragging down the popularity of MT use, and at present, the task assignments for post-editing jobs are 15 highly in demand and they are labelled as MTPE, a recognised acronym for machine translation post-editing, in the professional market. In addition, post-editing needs of clients vary greatly, which causes a different process for each one of them. Some clients, for instance, may only want the post-editor to correct obvious grammatical and syntactic errors, typos, misspellings and punctuation issues and not even make any preferential changes. There could be many reasons for such lower expectations from post-editors, which may include that the client need the translation desperately in a very short time, the translated text not to be published, or the pricing to be quite low for each party (or at least for the post-editor) and they could not expect anyone to spend more time on the assignment. On the other hand, some other clients may actually require that the post-editor correct all the mistakes, such as omissions, additions, mistranslations, punctuation problems and typos, by thoroughly reviewing the raw output and even making preferential changes where they deem appropriate to improve fluency and readability. Such assignment instructions suggest that the text to be translated (and possibly the client himself/herself) is quite important, and may be published, or distributed internally or externally. In addition to these, there are certain companies that are developing their own internal machine translation systems. These companies aim to feed their system with the best possible translations approved/corrected by good human translators to improve their system until it is finally perfect and does not probably need any human involvement one day. Accordingly, because these are internally developed systems, the domain is inherently restricted (such as biochemistry, pharmaceutics, pharmacovigilance), and therefore, these companies generally have glossaries prepared and reviewed beforehand and require that all post-editors to follow the glossary during post-editing without any exception. They provide the translation service provider with these glossaries or termbases along with very detailed and often long MTPE instructions. Because they invest time and money in these systems, they want consistent and quite uniform translations among all possible providers (there may be many different individuals/teams working for the same company, generally to ensure that someone would be available for any assignment). Although these glossaries or termbases are serving well for the purposes of the companies, the translation process is not that simple and almost no single term could be translated only in one way. These restrictions posed on translators may cause frustration in translators/post-editors. For 16 they have to obey these pre-approved translations of the terms despite they know that the imposed translations are not appropriate for the context. Accordingly, as Marshman (2014, p. 382) states, the translators may feel restricted as they are required to use matches retrieved from other resources. These matches, in such a case, may be from the database fed with other translations and then used to create the machine-translated text. The translators may also feel that their expertise and translatorial decisions are underestimated and not valued. Notwithstanding, the companies are quite meticulous in their own way that they often require two different post-editors to work on an assignment. While the first post-editor acts more like a regular translator and aims to create a very good quality translation as if it is performed by a human translator from scratch, the second post-editor ensures that the post-edited translation is accurate, consistent and error-free in every aspect. As a professional translator working in the market over 10 years and dealing with such new ways of translation practice, this is a process that I would like to call “dual post-editing”. While it may be argued that the second editor revises a human translation, not a machine translation, it is actually not the case. Because such assignments generally instruct the editors not to make any preferential or unnecessary changes, preventing them from exhibiting their own style, leaving behind a flat translation predominantly produced by a machine. While the raw output is not very close to the version after first post-editing, the clients assume it is. Besides, although these types of dual post-editing assignments underline the importance of human intervention, they also may mean that the more the human translators feed the system with more quality inputs the higher the risk of replacement by the system. However, it seems that this is the reality of the translation market and any translator who would like to exist and have a share in this market should adapt to these novel ways of working. In addition to the aforementioned processes, as of its development, MT has been used by and for different groups of people for different purposes. Nonetheless, the early assumptions on MT were not feasible and expectations were impractically high for MT systems, and as Hutchins (2001, p. 6) claims, the failure to recognise that the needs of users vary and each specific system should be intended for use by a certain group of users has fed the misconceptions about these technologies and their effects on translators. Therefore, it would be appropriate to elaborate on these different reasons 17 for the use of MT. Although Forcada (2010) lists two reasons of MT use (assimilation and dissemination), Hutchins’ listing with four purposes (actually assimilation, dissemination, interchange and database access in 2003 and 2005, assimilation, dissemination and communication in 2007) seems more comprehensive. Hutchins (2003, 2005 and 2007) explains “machine translation for dissemination” as the generation of translations that are in publishable quality and clarifies that these texts are not necessarily published but they are of that level of quality. Because these need to be proper and accurate, the process often involves a professional human translator, which means that the text is subject to all or some of the processes of pre-editing, post-editing the crude and sometimes lexically illogical output, employing a controlled language or using a system restricted to a single domain of a certain subject. He puts forward that MT is used for “assimilation” purposes when no human revision is required and only the essence of the message, the general idea of the text is sufficient for the recipient who is mostly a non-specialist layperson. For instance, the results of Vieira et al.’s study (2023) with 1,200 UK residents indicate such a use of MT for the need to read a document in a foreign language. In addition to these, people may use MT for communication purposes in social interchange if the quality of the text is not important or primary and the objective is to translate an electronic e-mail or another type of correspondence and to basically receive and deliver information or a message. This kind of MT use corresponds to what Hutchins (2003 and 2005) also categorises as interchange. On the other hand, database access includes resorting to MT for obtaining information written in a language that the user does not know well, such as searching and browsing the web pages. Forcada’s (2010, p. 217) description of assimilation encompasses database access and dissemination is the same as Hutchins’ explanation. It may be safely claimed that all of these purposes are applicable and valid today and many people with different backgrounds and needs use MT for one or all of these purposes. Not surprisingly, machine translation has not been running around in circles over the decades, and it made a remarkable progress with state-of-the-art technological advancements. Perceptibly, it started with code-breaking machines, direct approach and relatively simple rule-based MT and evolved into the most recent neural MT (NMT), 18 involving various other important developments in between (i.e. example-based and statistical MT). MT systems are designed for one single language pair (bilingual) or for multiple language pairs (multilingual) and while bilingual systems are generally designed to run unidirectionally such as from English into French, multilingual systems are rather bidirectional (Hutchins, 1995, p. 431-432). These systems in general have been operating based on three separate approaches with different levels of sophistication, namely direct approach (the oldest), interlingua approach and transfer approach (the last two are involved in rule-based approaches) (Hutchins, 1995, p. 432). Furthermore, corpus-based approaches should also be added to this list as the most sophisticated ones (at least for now) and including the example-based and statistical-based methods. These approaches are illustrated in Figure 1: (Quah, 2006, p. 68) Figure 1: MT Architectures 1.2.1.1. Direct approach Direct approach with a primitive bilingual and unidirectional system was the first approach which was developed by engineers and mathematicians for languages with close grammatical structures as it could not handle metaphors, complicated and long sentences and consider semantics, that’s why it could not be any successful for distinct languages with its rather basic syntactic analysis (Quah, 2006, p. 69-70). It was mainly based on word-for-word translation approach, replacing source text words with target MT Approaches Direct Translation (1st Generation) Rule-based (2nd Generation) Transfer Interlingua Corpus-based (3rd Generation) Example- based Statistical- based 19 text words while re-ordering the words as necessary; however, the system was far too simple and was generating unreliable and poor results that could not address professional translation needs (ibid., p. 70). CULT, Weidner, METEO, and the old Systran could be given as examples to early machine translation systems employing direct approach (ibid.). 1.2.1.2. Rule-based approaches Dominant until the 1990s, rule-based machine translation (RBMT) is based on “morphological, syntactic, semantic, and contextual knowledge about both the source and the target languages, respectively, and the connections between them to perform the translation task” (Shiwen and Xiaojing, 2015, p. 186). In other words, it depends on the assumption that translation involves the analysis and representation of SL text “meaning” to generate its equivalent in TL (Quah, 2006, p. 71). RBMT includes two approaches, namely interlingua and transfer approaches, which have intermediate or abstract representations. 1.2.1.2.1. Interlingua approach This approach depends on the analysis and conversion of an SL text into its interlingua representation which is universal or independent of the language (appropriate for multilingual translation needs) (Cheragui, 2012; Hutchins, 1995; Quah 2006). Texts are translated into other languages from these representations, which means this approach is not direct but has two steps as from SL to the interlingua (analysis stage) and then from the interlingua to TL (generation or synthesis stage) (Cheragui, 2012, p. 164; Hutchins, 1995, p. 432). Although it is economic for multilingual environments as it does not require much work to add a new language to the system (Hutchins, 1995, p. 432), the aim of defining a universal, language-neutral and common representation that would work for any language was problematic in practice (Quah, 2006, p. 73). The interlingua-driven systems, the main investigation centre of which was Leningrad University (Hutchins, 1995, p. 436), were applied as the main approach for the development of “knowledge-based machine translation” by Carnegie Mellon University, which was based on the idea that translation involves more than linguistic knowledge (Quah, 2006, p. 72). This new approach was an improved version of the 20 interlingua system, infused with pragmatic and semantic knowledge and utilising a parser to translate an SL text into a “semantic” representation and then generating a TL text based on this representation (ibid., p. 72-73). KANTOO (newer version of KANT) by Carnegie Mellon University, Mikrokosmos, Pivot, HICATS could be given as examples of this more advanced knowledge-based system, and DLT and Rosetta could be given as examples of the essentially interlingua- and linguistics- oriented systems (ibid., p. 73). 1.2.1.2.2. Transfer approach The transfer approach operates in three stages rather than two in sequential order: analysis, transfer, and synthesis or generation. In the first stage, an SL text is converted into an abstract SL representation followed by the second stage where the aim is to transfer the SL representation into TL representation (Quah, 2006, p. 73). And in the final stage, the system generates a TL text. This approach, unlike the interlingua approach, runs different dictionaries and transfer models at each stage (SL dictionary, a bilingual dictionary and a TL dictionary, respectively), and thus, it requires more work to add a specific language to this system, examples of which include Ariane, SUSY, METAL, Atlas-I and Duet (ibid., p. 73-75). 1.2.1.3. Corpus-based approaches Corpus-based approaches (or as Cheragui (2012, p. 165) and Forcada (2015, p. 161) note, data-driven approaches) became dominant and gained popularity in 1990s. Back then, there were two different methods, namely statistical- and example-based methods, however, 2010s saw the introduction of neural machine translation, the most recent type of corpus-based approaches. As their names indicate, all kinds of corpus- based approaches leverage a corpus that contains bilingual, parallel texts as reference, and the system uses these references to generate translations for new sentences. In these approaches, unlike rule-based methods, pre-determined linguistic rules are not relied on for the analysis of texts or for matching new strings with the previously translated equivalents in order to produce new ones (Quah, 2006, p. 77). 1.2.1.3.1. Statistical-based approach 21 Although there were efforts towards statistical machine translation (SMT), they were not successful until the 1990s, and the IBM’s introduction of Bayes’ theorem in their Candide system was the first impressive improvement for SMT in experimental environment (Quah, 2006, p. 77). This approach was based on the idea that “a translation can be modelled with a statistical process” (ibid.). It is obvious that a large number of aligned strings are needed for this approach to work since as Trujillo (1999, p. 211) explains, “Translation […] requires a method for (a) computing the probability of a string being the translation of an SL string; (b) computing the probability of a TL string being a valid TL sentence, (c) a technique to search for the TL string which maximizes these probabilities. However, as Quah (2006, p. 77) underlines, in most SMT systems, the translation model is based on single-words, which means there is only one TL word for a single SL word. As translation is too complicated to employ a word-for-word approach, and because this approach does not take the surrounding words (i.e. context) into consideration, pure SMT has disadvantages in practice. Notwithstanding, the phrase-based SMT operating on a “joint probability model” proposed by Marcu and Wong (2002), which “automatically learns word and phrase equivalents from bilingual corpora” and the related translation model and decoding algorithm proposed by Koehn et al. (2003) definitely contributed to the success of SMT, showing that these phrase-based models outperform the word-by-word systems, such as IBM Model 4. 1.2.1.3.2. Example-based approach Example-based machine translation (EBMT) was first proposed by Nagao (1984) arguing that then available MT systems need improvement, and accordingly, he suggested a model based on the “language processing in the human brain” operating on analogy principle, and founded his model on the idea that for translating a sentence, humans decompose a sentence into fragments and translates these fragments to generate a long sentence in the end, rather than performing linguistic analysis (Nagao, 1984, p. 175). He claims that this analogy-based system needs a huge corpus with example translations and is required to notice similarities and differences between the new segment and the previously translated ones. While there could be only one similar SL sentence in the corpus, there may also more than one reference sentence and in that 22 case, the model matches the similar parts of the new segment with the stored SL segments and combines the fragmental phrases to produce the translation of the new segment. Although this model reminds the mechanism of translation memories (TM) in computer-aided tools, TMs can only extract a fuzzy or full match from one single previously stored translation, which is generally the most up-to-date one (but this may change based on user settings), and they cannot combine two or more matched sentences. As Quah (2006, p. 83) underlines, this model may fail based on the scope of the corpus as there may be no close matches for any segment, and also its success is controversial for agglutinative languages, such as Turkish, that would make matching process more complicated. It may also be argued that for minority languages or for languages that do not have as many parallel texts as the major ones, these models can dramatically fail depending on the lack of sufficient corpora, and rule-based approach may be the way to obtain better quality. However, recent improvements may have found a solution to this problem. 1.2.1.3.3. Neural-based approach One of the scholars who investigated ways to improve SMT systems was Schwenk (2012), and in his paper, he introduced a model that could be integrated in phrase-based SMT system and uses “neural networks to directly learn the translation probability of phrase pairs using continuous representations.” As a result of his experiment, it was concluded that the model could “infer meaningful translation probabilities for phrase pairs not seen in the training data” (2012, p. 1071) and because the model has the ability to provide translations not in the training data, he argues that: An interesting extension of this idea is to use large amounts of monolingual data to pre-train the projections of the source words onto the continuous space. The neural network could learn from the monolingual data that words are synonyms since they often appear in similar contexts. […] This could be also interesting when translating from a morphologically rich language into English since many verb forms actually translate into the same English word. (Schwenk, 2012, p. 1078) Another salient work on neural MT was performed by Bahdanau et al. (2014). Building upon the phrase-based models (Koehn et al, 2003; Marcu and Wong, 2002), a novel neural network model was introduced by Bahdanau et al. (2014). The scholars 23 proposed RNN Encoder-Decoder that “consists of two recurrent neural networks (RNN) that act as an encoder and a decoder pair” with a “novel hidden unit” (2014, p. 1724): We train the model to learn the translation probability of an English phrase to a corresponding French phrase. The model is then used as a part of a standard phrase-based SMT system by scoring each phrase pair in the phrase table. The empirical evaluation reveals that this approach of scoring phrase pairs with an RNN Encoder–Decoder improves the translation performance. (Bahdanau et al., 2014, p. 1724) Such scholars achieved more success in terms of quality output generated by machine translation systems as they managed to develop machines that could operate as human brain in a similar way. Given that this is what is attempted with neural networks3, it is not surprising that machine translation has progressed tremendously in the last decade. However, NMT4 is still not without disadvantages and drawbacks. In their experiment, Koehn and Knowles (2017) compare the conventional SMT with NMT and list six areas for improvement where NMT: (i) has lower quality “out of domain”, (ii) has lower quality for low-resource languages, (iii) has lower quality for rare words in highly-inflected languages, (iv) has lower quality when translating long sentences, (v) has weaknesses in terms of word alignment model, and (vi) translation quality gets worse when a larger search space is in question for beam search decoding. Although the aforementioned categorisation for machine translation systems may be useful theoretically, the systems are now based on multiple approaches rather than just one approach. Hutchins (1995, p. 432) highlights that this “third generation” of hybrid systems are quite common where rule-based approaches are integrated with example- based and statistical-based approaches, transfer approaches are fused with interlingual components or interlingua and transfer systems are used together, etc. Quah (2006, p. 84) justifies this hybridity with an attempt to combine the best aspects of different systems, trying to mitigate the disadvantages of each one of them. 1.2.2. Online MT systems 3 A neural network may be defined as a system “that simulate intelligence on the computer to imitate the way a human brain works” (Quah, 2006, p. 36). 4 For an additional and detailed explanation and definitions of terms in NMT jargon, see Forcada, M. L. (2017). Making sense of neural machine translation. Translation spaces, 6(2), 291-309. 24 In addition to all of these offline models, both free and paid online systems over the years have also been introduced, including neural MT systems. Quah (2006, p. 85-86) states that the aim for the development of MT systems was to translate formally written natural languages, but globalisation and increased use of the Internet created new requirements for translation, such as translating e-mails, and webpages. While globalisation and the resulting increased volume of translation may be claimed to trigger the development of “client-server” MT systems (ibid.) designed for professional use, the Internet, in particular, may be argued to have caused an increased need for translating informal content by non-translators. In such a client-server system, a language service provider or an organization hosts the system on a server and the client, a translation company, organisation or a translator, can access the machine translation system remotely and via the Internet with specific credentials (username and password). Furthermore, Olohan (2021, p. 105) puts forward that many technology companies help the language service providers (LSPs) on certain aspects of MT. Firstly, they offer the service of developing MT engines for LSPs, which frees LSPs from dealing with technical matters, and secondly, and more importantly, they offer remote access to their online platforms, which eliminates the need to buy certain hardware or install and set-up the relevant software as well as the need to ensure and maintain high data confidentiality and security levels. However, Olohan (ibid.) warns that the companies may provide the MT engines, but they cannot produce the translated segments to feed the engines. The responsibility of providing an ideally huge and good quality bilingual corpora lies with the LSPs and that is a significant factor in obtaining good quality MT outputs. This is a good indicator that shows how the workflow begins to change from the top. Accordingly, this change will eventually affect the working conditions of translators as they get jobs from such LSPs, which may be argued to show an increasing trend for MTPE assignments. Olohan (ibid.) supports this argument and states that MT use by LSPs is mainly as “human-in-the-loop systems” that indicates the involvement of human translators mostly as post-editors. These systems can also be accessed by non-translators but generally they are not, as these systems are not intended for this group of end-users. Such services are paid due to the professional nature of use. On the other hand, free online MT systems are usually chosen by non-translators who need to quickly grasp the main idea of a text, understand 25 the features of and buy a product sold at a foreign-language website, or communicate with their colleagues and clients, and so forth. For such users, the quality of the output is not primary and could be ignored to the extent that the main goal of the user is achieved, such as learning the technical specifications of a device that they would like to purchase. Free MT systems include Google Translate (GT), Bing (Microsoft Translator), Babelfish, WorldLingo, and Promt Online among others. Hutchins (2001, p. 18-19) argues that the availability and popularity of such free online MT systems will draw the attention of the public to the significance of translation in international communication, and people will start to think more of the profession of translation as they see the poor-quality raw output, resulting in even more demand for good quality human translations. Furthermore, he adds that neither MT nor automation will cause any threat to the profession, on the contrary, it will improve the working conditions and increase the volume and amount of business (ibid., p. 19). Besides, it should be noted that the MT system developers need human translators to create a huge database of parallel texts, so they need to accommodate the needs of translators to remain powerful in the market. However, there is a maturation stage for each technological development (Beaudry et al., 2016, p. 201), and at that point, the need for human translators may also disappear. Moreover, under the present market conditions, the use of MT is still often required by the client companies. However, the improved quality and PC-based MT systems are not considered a valid option by most translators as Quah supports (2006, p. 91). This may be due to the fear of replacement, not trusting the output quality, spending more cognitive and technical effort for post-editing, confidentiality issues which may arise in case of using free online systems, the high price of licensed MT systems, or just out of habit. As opposed to avoiding MT, translators embrace the concept of the CAT systems (o