How does post-editing – the act of revising machine-translated output – affect the feminist voice in translation? My PhD research into contemporary Russian poetry and machine translation has led to me reconsider the relationship between technology and feminist writing practices. Instead of seeing machine translation as something to be overcome or corrected, I’ve begun exploring how its very limitations can be productively engaged with to further a feminist translation practice. Machine-translating the feminist source texts of my project presents challenges on both a linguistic (for example, gendered language, wordplay and case systems) and a cultural level: the increasingly restrictive gender politics of Russia is apparent in the poets’ confrontations of social taboos around female sexuality and gender norms. Such challenges can, however, become springboards for feminist innovation in the post-editing process.
Machine-translation systems, trained on vast corpora of (predominantly male-authored) texts, tend to reflect, reproduce and perpetuate societal beliefs and biases. I’ve discovered in my research that when these systems consistently default to masculine pronouns, medicalise female bodily experience or struggle with non-binary terms, they inadvertently make visible the structural biases embedded within language. Rather than just ‘correcting’ these translations, feminist post-editors can strategically highlight them to show how language can function to uphold patriarchal structures. These aren’t necessarily ‘errors’ to be fixed, but rather tangible evidence of how gender operates within certain discourses. The post-editing process creates a relationship of constant renegotiation and repositioning between the translator, source text and machine-translated output. I’ve found that the tension between my translation instincts and the system’s suggestions allows me to question and investigate my own internalised translation biases. For example, what do I consider ‘natural’, ‘fluent’ English to be? What has shaped this belief?
Engaging with machine-translation systems in such a way challenges wider translation norms: the ideal of a translator who perfectly comprehends and controls the translation process is usurped by a model of (machine-)translation as collaboration, negotiation and partial perspective. When post-editing, my acknowledgement of each machine-translated suggestion (even if I ultimately reject it!) embodies a translation approach that recognises its own limitations and uncertainties, and values situated knowledge over perceived perfection and objectivity. In my work, I see the mistakes, glitches and inconsistencies of the machine-translation system as invitations to critically examine my own practices and assumptions as a feminist translator. The post-editing process contains possibilities for new, unexpected forms of linguistic and textual intervention; the machine takes on the role of an unpredictable collaborator in developing more reflexive and ethical translation methodologies.
Such technological entanglement indeed changes me as a translator. In seeing the potential value in machine ‘error’, I’ve come to develop a deeper understanding of my own unconscious translation habits. The reality of working with machine-translation systems is, of course, often more pragmatic and constrained than the theoretical frameworks of my research imply. But in imagining what feminist post-editing could be, translators can become more aware of their own biases and assumptions, acknowledging and embracing the multitude of possibilities and uncertainties that translation offers up for exploration.
Matilda Hicklin, University of Bristol