A communicative approach to exploring Machine Interpreting in conference settings

What is M.INTerpreting?

In recent years, machine interpreting (speech translation) has gained momentum thanks to the swift advances in artificial intelligence. Due to dramatic improvements in the quality of interlingual translation and automatic transcription, academic and commercial applications have been developed for use in non-professional and semi-professional environments, and they are expected to enter the professional market at some point in the future. National and international institutions, for example, are currently looking at the implementation of such technologies to support multilingual communication.

Empirical investigation on machine interpreting has so far mostly been conducted within the framework of computer science, for instance with the goal of improving the current widely used cascading approach to speech-to-speech translation, or of exploring the potential of end-to-end systems. So far, investigations have been technology-centred and evaluation has been conducted with the goal of measuring improvements over time and developments cycles.

What is still missing is a communicative approach to the analysis and evaluation of machine interpreting that considers, among others:

  • The use of machine interpreting systems in real interactions between humans
  • The perception of the output by the communication actors depending on the system's approach
  • The adequacy and efficacy of communication mediated by MI systems
  • The implication of its use in high-risk settings, ethics, etc.
  • The influence of MI on the environment for professional interpreters.

These are the areas that we want to cover with our project.

Within our project we are developing a baseline translation system (cascading system).

The project is funded by the Mainzer Inneruniversitäre Forschungsförderung

Goals of the project


Development of a linguistic framework for a communication-oriented analysis of machine interpreting


Creating a corpus and a methodology to test MI from a communicative prospective


Analysis of fundamental phenomena of machine interpreting (completeness, adequacy, reception, etc.)


Investigating potentials and limitations of machine interpreting in complex contexts and authentic institutional communicative situations (parliamentary debates, bilateral meetings, etc.).

Ethics and Risks

Overcome the limited understanding of the nature of risks related to the uninformed use of MI and of the wider implications of using this technology (for example in terms of ethical issues)

Comparison Machine and Human interpreting

Evaluating MI not only against perfect standards but also against real human interpretation performed on the same dataset

Target group


Our contribution

While first attempts at a communication-oriented investigation of MI in the field of dialogue interpreting are already available, the application of MI to conference interpreting remains completely unexplored. Our project aims at filling this research gap by addressing the phenomenon in the field of conference interpreting (both simultaneous and consecutive) from a communicative perspective.

We want to achieve this by combining the humanities with computer science.


Feel free to get in touch with us if you have any questions or suggestions.

Dr. phil. Claudio Fantinuoli

Claudio Fantinuoli is a Senior Lecturer and Researcher at the University of Mainz/Germersheim. He is the founder of the CAI tool InterpretBank and head of the NLP innovation team at Kudo Inc. He teaches an introductory class in Speech-to-Text translation and augmented human interpretation at the Post Graduate Center of the University of Vienna and collaborates with the Speech-To-Text Unit of the European Parliament as an external expert.

Claudio is interested in multilingual communication and the diverse approach to speech translation between humans and machines as well as in how new technologies, especially AI, can support human interpreters. To gain insights into these questions, he develops and studies computational systems to automatically translate speeches or to assist interpreters in doing so. The ultimate goal of Claudio's research is to empower human interpreters in their activity and to bring about a more precise characterization of what is unique in human interpretation.

Claudio Fantinuoli
Interpretation, NLP, Artificial intelligence

Bianca Prandi MA

Bianca Prandi holds a Master’s Degree in Conference Interpreting from the University of Bologna/Forlì. She is currently a Ph.D. candidate and a research associate at the University of Mainz/Germersheim. She is the co-founder of Interpremy, a research-based online training project for conference interpreters and interpreting students, and collaborates with the Postgraduate Centre of the University of Vienna as a lecturer in Computer-Assisted Interpreting and Remote Simultaneous Interpreting.

Her current research project focuses on how Computer-Assisted Interpreting tools can enhance interpreters’ performance and on exploring the impact of their integration in interpreters’ workflow from a cognitive perspective. Her research interests include Computer-Assisted Interpreting, cognition in spoken multilingual communication, and Natural Language Processing. With her research, she hopes to dispel common misconceptions about technology-assisted interpreting and to empower interpreters through technology.

Bianca Prandi