Sara Mills defines a post-feminist text analysis as “one which is able to see that there are, within the parameters of the textual and discursive constraints, multiple interpretations of terms and discourses as a whole.” (Mills, 1998) Mills’ approach takes into account the impact of feminism and offers a more nuanced perspective compared to earlier analyses that focused on overt sexism and discrimination. Mills references Toolan’s argument that analysis should move away from “easy examples of sexism and racism” and to “subtler and hence more insidious discriminatory and exclusionary discourses that abound.”

Mills’ approach can be useful in evaluating, generating, and re-envisioning the machine-aided methods of text analysis that are becoming increasingly prevalent. In Lauren Klein’s book Data Feminism (2020), Klein offers a similar critique by comparing approaches to data by Facebook and the Make the Breast Pump Not Suck Hackathon. While Facebook expanded gender categories, their policies did little to protect and instead actively imperiled “the safety of some of the platform’s most marginalized users.” On the other hand, used reductive data categories to achieve goals of inclusion and equality. Klein’s examples demonstrate that a more nuanced approach to data and text analysis is necessary to to identify and address the subtler forms of discrimination and exclusion that can still occur.

In the field of translation, machine-assisted approaches are becoming more common. Although machine-assistance can improve efficiency and fidelity, it introduces another layer of complexity to the process of translation. Mills’ approach could aid in examining not only the product of translation but also the underlying systems and processes, including machine-assisted translation. While Taivalksoki-Shilov (2019) addresses some of the ethical concerns regarding machine-assisted translation such as the quality of translation and preservation of the author’s voice, other considerations are absent. Mills’ approach to textual analysis provides a useful framework for considering a broader range of perspectives and contextualizing editorial and linguistic choices within the cultural norms and ideologies of the source text.

Klein, Laura. What Gets Counted Counts. Data Feminism. MIT Press, 2020.

Mills, Sara. Post-feminist Text Analysis. Language and Literature, 1998, p. 241.

Taivalkoski-Shilov, Kristiina. Ethical issues regarding machine(-assisted) translation of literary texts. Perspectives 27.5 (2019): 689-703. DOI: 10.1080/0907676X.2018.1520907.

If, as Sara Ahmed writes, feminism is “a building project” (Living a Feminist Life, 14), then a critical point of feminist intervention in the textual analysis pipeline is in the selection, handling, and use of textual data. The text tokens — and the methods we use to parse, tag, and untangle them — are the nails and hammers, respectively, in that building project.

Every stage of identifying and preparing data (i.e., prior to its analysis) offers opportunities for feminist intervention. When working with an existing dataset (a readymade), we as researchers can interrogate its provenance, genealogy, potential for misuse, and more. Asking these questions can clue us into the assumptions that shaped its collection. 

If collecting data ourselves, we should look to domain experts to help us circumvent our own biases and situate knowledge within its local context — otherwise, we risk harming the communities the data claims to represent. When cleaning data, we have an opportunity to preserve difference and nuance by avoiding the tendency to aggressively standardize data to make it “easier” to analyze. 

We also have opportunities to rewrite or redistribute data in ways that actively oppose oppression — gender-based or otherwise — by revising or adding metadata, documenting and explaining datasets to promote more informed use, and even constructing alternative datasets that subvert dominant ideologies. These actions can not only bring our research efforts in line with a feminist aim, but also create the conditions for other researchers to do the same.