67 Spring 2026 Topic 5, the Final Solution topic (marked by the brown bar in figure 2), is dominated by the defendant’s testimony and, as we shall see, by the Court’s judgment. Its vocabulary is that of bureaucratic administration: the names of Nazi officers in the chain of command (Himmler, Heydrich, Müller), the official designation of the plan and its administrative structures (Department, Order, IVB4), the geographic and organizational coordinates of its execution. It is the vocabulary of someone who knew the plan from above, who saw the system whole. Topic 6, the Death Camp topic (marked by the pink bar in figure 2), is dominated by survivor testimony. Its vocabulary is almost the mirror opposite: Barrack. Gas. We were. Happened. Remember. Picture. Inside. These are not the words of someone describing a system. They are the words of someone describing an experience that arrived without explanation, without context, without the administrative frame that would have made it legible. If Topic 5 provides what perpetrators’ bureaucratic view of the Final Solution, Topic 6 exposes what it looked like from within: an inexplicable and total reality or annihilation. The two topics share, across their ten top terms, only two words: “gas” and “Auschwitz.” The same historical event, the murder of millions of people, produced two almost entirely non-overlapping vocabularies depending on the position of the speaker. This is not merely a difference of register or formality. It is evidence of what might be called an epistemic divide: the incommensurable gap between the perpetrator’s organizational knowledge of destruction and the victim’s embodied experience of it. The computational analysis makes this divide visible with a precision that close reading of selected testimonies cannot achieve, because it holds both vocabularies simultaneously across the full corpus. The sharpest test of testimony’s legal influence, however, is not found in the trial’s breadth but in its culmination. What the topic analysis makes visible is where the trial ultimately landed, in whose language the Court chose to render its verdict. Topic 5 (Final Solution), with its lexicon of orders, departments, and the administrative machinery of annihilation, registers the third highest mean proportion in the Judgment, suggesting that when the Court moved to render its legal conclusions, it reached instinctively for the language of perpetration. Topic 6 (Death Camp), the vocabulary of those who were there, barracks, gas, the first-person witness, is among the lowest-proportioned topics in the Judgment, present at the margins of the Court’s legal reasoning despite dominating the testimony phase that preceded it. The prosecution and the Court, in other words, concluded the trial in the language of the accused. The experiential vocabulary of annihilation, barracks, gas chambers, the first-person plural of those who were there, did not travel from the testimony phase into the legal reckoning. The bureaucratic vocabulary of administration, departments, orders, the organizational hierarchy of destruction, did. This finding sits in uncomfortable proximity to a declaration the Court itself had made. In the Judgment, the justices described survivor testimonies as “by-products” of the trial, valuable to historians but peripheral to the legal case. The computational analysis provides empirical corroboration for precisely this self-assessment: the topical profile of the Judgment aligns far more closely with the defendant's discourse than with the survivors. Examining which topics are most associated with each speaker group, the Court and the victims share only a single dominant overlapping topic (Deportations), while the Court and the defendant share two (Nazi Bureaucracy and Final Solution). And even Deportations, ostensibly a survivor-centered theme, is one in which Eichmann’s direct role and administrative responsibility, his orchestration of the deportation trains, was centrally at issue. The trial remembered as the paradigmatic case for victim-centered justice produced, in its formal legal output, a judgment written in the language of the perpetrator it convicted. V. What the Algorithm Cannot Hear — And why that Matters No method is without its limits, and the limits of computational text analysis are worth naming precisely because they are themselves revealing. The topic model found no statistical trace of the Zionist meta-narrative that critics of the prosecution, Arendt most prominently, accused Hausner of imposing on the proceedings. Terms relating to heroism, resistance, immigration to Israel, national redemption: none of these formed a coherent statistical cluster in the corpus. This absence does not mean such themes were never invoked; it suggests they did not constitute a dominant or statistically coherent strand of the trial’s linguistic fabric. Whatever rhetorical resonance the Zionist frame had at the level of individual speeches or moments it did not organize the trial’s overall discourse in the way the cultural criticism has implied.
RkJQdWJsaXNoZXIy MjgzNzA=