Models of the Translation Process and the Free Energy Principle.

Area of Science:

• Cognitive Science
• Linguistics
• Computational Neuroscience

Background:

• Translation process research (TPR) has numerous models explaining human translation.
• Existing models often lack a unified theoretical framework for cognitive and pragmatic aspects.

Purpose of the Study:

• To extend the monitor model by incorporating relevance theory (RT) and the free energy principle (FEP).
• To provide a generative model for elucidating translational behavior using FEP and active inference.
• To formalize key concepts of TPR, RT, and the monitor model within a unified mathematical framework.

Main Methods:

• Adopting the free energy principle (FEP) as a generative model for translational behavior.
• Analyzing behavioral data based on translation units (TUs), measuring translation effort and effects.
• Mapping FEP concepts (free energy minimization, active inference) onto the translation process.
• Clustering TUs into translation states (steady state, orientation, hesitation) and sequences into policies.

Main Results:

• Demonstrated compatibility between free energy minimization and the concept of relevance in RT.
• Showcased how sequences of translation states form policies that reduce expected free energy.
• Formalized essential concepts of the monitor model and RT as deep temporal generative models.

Conclusions:

• The FEP offers a robust mathematical framework for understanding human translation as goal-directed, information-seeking behavior.
• This integrated approach provides a unified account of cognitive, pragmatic, and behavioral aspects of translation.
• The model supports both representationalist and non-representationalist interpretations of translation processes.