This book is a contribution to the research community towards thinking and reflecting on what Responsible Machine Translation really means. It was conceived as an open dialogue across disciplines, from philosophy to law, with the ultimate goal of providing a wide spectrum of topics to reflect on. It covers aspects related to the development of Machine translation systems, as well as its use in different scenarios, and the societal impact that it may have. This text appeals to students and researchers in linguistics, translation, natural language processing, philosophy, and law as well as professionals working in these fields.
Author(s): Helena Moniz, Carla Parra Escartín
Series: Machine Translation: Technologies and Applications, 4
Publisher: Springer
Year: 2023
Language: English
Pages: 241
City: Cham
Foreword
Acknowledgements
Contents
Editors and Contributors
Abbreviations
Chapter 1: Introduction
1.1 Responsible Artificial Intelligence
1.2 Towards Responsible Machine Translation
1.3 Structure of the Book
1.4 Avenues for Future Work
References
Part I: Responsible Machine Translation: Ethical, Philosophical and Legal Aspects
Chapter 2: Prolegomenon to Contemporary Ethics of Machine Translation
2.1 Introduction
2.2 Translation and Ethics
2.3 What Machines Do To Translation
2.4 Ethical Questions of MT
2.5 Conclusion
References
Chapter 3: The Ethics of Machine Translation
3.1 Introduction
3.2 Language and the Problem of Meaning
3.3 From Meaning Construction to Meaning Consumption
3.4 Semantic Mismatches in Meaning Formulation
3.4.1 Digital Language Misrepresentation
3.4.2 Digital Language Pluralism and Reviews
3.4.3 Multilingual Cross Sectoral Data Access and Sustainability
3.5 The Ethics of Semantic Design
3.5.1 Linking Lexical Knowledge at the Level of Meaning
3.5.2 Discovery Search and Display
3.6 Conclusion
References
Chapter 4: Licensing and Usage Rights of Language Data in Machine Translation
4.1 Introduction
4.2 Machine Translation Relies on Data
4.2.1 Rule-Based Machine Translation
4.2.2 Corpus-Based Machine Translation
4.2.3 Hybrid Systems
4.3 Translation Data, Usage Rights, and Licensing
4.3.1 Linguistic Resources for Machine Translation
4.3.1.1 Licensing of Linguistic Resources
4.3.2 Sentence-Aligned Parallel Corpora
4.3.2.1 Sentence-Aligned Parallel Corpora Published by Their Owners
Corpora Published by Public Administrations
Corpora Expressly Created for Open Software or Other Projects
Other Sentence-Aligned Corpora
4.3.2.2 Web-Crawled Sentence-Aligned Parallel Corpora
Corpora from Non-governmental Organisations
Corpora from Religious Sites
Corpora from Public Administrations
4.3.3 Licensing and Usage Rights of Sentence-Aligned Parallel Corpora
4.3.3.1 Corpora Published by Their Owners
4.3.3.2 Web-Crawled Corpora
Automatic Copyright
Copyright as Protection
A Nonliteral Approach to Copyright
Legal Exceptions to Allow for Text Mining
A Partial Workaround: `Deferred Crawling´
Translators as Authors
Value Added and Changes in the Profession
Claims for Compensation for Subsequent Use
4.4 Concluding Remarks
References
Chapter 5: Authorship and Rights Ownership in the Machine Translation Era
5.1 Translation as an Intellectual Work
5.2 Translation by Means of Machines
5.3 Translation Authorship with Current Systems
5.4 The Blurring of Authorship in Advanced Systems
5.5 The Pretended ``Electronic Personality´´ of an Intelligent System
5.6 Fully Automated Translation
5.6.1 Approach
5.6.2 The System Programmer
5.6.3 The System User
5.6.4 Legal Solutions Outside the Copyright Field
5.7 Use of Translation-Associated Results
5.7.1 Personal Data Associated with Translation
5.7.2 Technical Data Suitable for Training Neural Networks
5.7.3 Rights Over Linguistic Resources
5.8 Intellectual Property Regulation and Corpora Extracted from Translation Memories
5.9 Conclusions and Main Takeaways
References
Part II: Responsible Machine Translation from the End-User Perspective
Chapter 6: The Ethics of Machine Translation Post-editing in the Translation Ecosystem
6.1 Introduction
6.2 MTPE in the Translation Ecosystem
6.2.1 Cooperation Partners in MTPE
6.2.2 Social Factors in MTPE
6.2.3 Artefacts in MTPE
6.2.4 Psychological Factors in MTPE
6.3 Ethical Dilemmas in MTPE
6.3.1 Ethical Dilemma #1: The Post-Editor´s Status
6.3.2 Ethical Dilemma #2: The Post-Editor´s Commitment to Quality
6.3.3 Ethical Dilemma #3: Digital Ethics and the Post-Editor´s Responsibility
6.4 Concluding Remarks
References
Chapter 7: Ethics and Machine Translation: The End User Perspective
7.1 Introduction
7.2 Modelling MT Use Cases
7.3 The Voices of Users
7.3.1 First Experiment: Technical Environment
7.3.1.1 Did You Notice a Difference in the Language in MS Word When You Came Back from the Pause?
7.3.1.2 How Did You Find the Quality of the German/Spanish/Japanese Here?
7.3.1.3 How Did You Find the Language in this Menu, Dialog Box, Option?
7.3.1.4 How Did You Feel During this Task?
7.3.2 Second Experiment: Creative Environment
7.3.2.1 If You Realised That It Was a Translation, Can You Describe How Did You Conclude This?
7.3.2.2 Were There Any Paragraphs or Sentences That Were Difficult to Understand? Can You Tell Us Which Ones?
7.3.2.3 Was There a Sentence or a Paragraph That You Especially Liked? Can You Tell Us Which One?
7.3.2.4 Do You Want to Make Any Other Comment?
MTPE
MT
7.4 Discussion on Ethical Implications
7.5 Conclusions on Ethical Implications
References
Chapter 8: Ethics, Automated Processes, Machine Translation, and Crises
8.1 Introduction
8.2 Automation Processes and Crisis Preparedness
8.3 Crowdsourcing Data, Mapping, and Translation Automation
8.4 From Local Cascading Crises to Global Events
8.5 From Monomodal to Multimedia Communication
8.6 Conclusions
References
Part III: Responsible Machine Translation: Societal Impact
Chapter 9: Gender and Age Bias in Commercial Machine Translation
9.1 Introduction
9.2 Machine Translation
9.3 Bias in Machine Translation
9.4 Gender and Age Bias in Commercial Machine Translation
9.4.1 Method
9.4.2 Data
9.4.2.1 Translation Data
9.4.2.2 Profile Prediction Data
9.4.3 Classifiers
9.4.4 Gender Bias
9.4.4.1 Translating into English
Using BERT: Language-Specific Mono-lingual Models
Using BERT: Multi-lingual Models
9.4.4.2 Translating from English
9.4.5 Age Bias
9.4.6 Discrepancies Between MT Systems
9.5 Discussion
References
Chapter 10: The Ecological Footprint of Neural Machine Translation Systems
10.1 Introduction
10.2 The Technological Shift(s) in MT
10.2.1 From Rule-Based to Neural MT
10.2.2 Why GPUs?
10.2.3 The More, the Better?
10.3 Related Work
10.3.1 Research
10.3.2 Tools
10.3.3 Workshop
10.4 Case Study: Empirical Evaluation of MT Systems
10.4.1 Hardware Setup
10.4.2 Machine Translation Systems
10.4.3 GPU Power Consumption
10.5 Power Consumption and CO2 Footprint
10.5.1 Train and Translation Times
10.5.2 Power Consumption and CO2 Emissions
10.5.3 The Impact
10.6 Optimizing at Inference Time Through Model Quantization
10.6.1 Quantization
10.6.2 Quality of Quantized Transformer Models
10.6.3 Energy Considerations for Quantized Transformer Models
10.7 Conclusions and Future Work
References
Chapter 11: Treating Speech as Personally Identifiable Information and Its Impact in Machine Translation
11.1 Introduction
11.2 Recent Progress in Speech Technologies
11.2.1 Computational Paralinguistics
11.2.2 Speaker Representation
11.2.3 Speech Recognition
11.2.4 Speech Synthesis
11.2.5 Voice Conversion
11.3 Privacy and Security Breaches in Speech Technologies
11.3.1 Growing Awareness
11.4 Voice Privacy
11.4.1 Anonymisation
11.4.2 Encryption
11.5 Speech-to-Speech Machine Translation
11.6 Conclusions
References
