Introduction
In the field of linguistics, "Information Status" describes the level of familiarity an audience has with a specific entity. When a person or organization is first introduced to the public, they are considered "hearer-new" and typically require a detailed description—for example, "Google, the company behind the popular Web search engine" in the early 2000s. Over time, as the entity becomes "common knowledge" or "hearer-old," these descriptions shrink until only a name is necessary: "Google."
While previous research has examined how these referring expressions (REs) change within a single document, a recent study provides the first longitudinal analysis of how this transition happens across decades of collective memory. By analyzing 20 years of The New York Times, researchers have mapped the linguistic "trajectory" of fame.
The Shrinking Descriptor
The primary indicator of an entity's transition to familiarity is the reduction in the length of its descriptors. The researchers defined the point of "acceptance" as the moment an entity's average descriptor length consistently stays below one word.
As shown in the figure above, the initial introduction of an entity is marked by heavy descriptive language, which drops precipitously within the first few years. However, the "speed" of this transition varies. While global figures like Bill Clinton reach "hearer-old" status in less than a month, other entities may take over a decade to achieve the same level of recognition, or may never reach it at all.
Methodology
To track these changes, the researchers developed a pipeline using the New York Times Annotated Corpus (1987–2007), which contains 1.8 million articles. They extracted noun phrases and separated them into a base expression (the name) and a descriptor (the additional information).
By linking mentions of over 52,000 unique people (PER) and organizations (ORG) across 20 years, they created "entity chains." These chains allowed them to analyze 61 different features—including syntax, frequency of mentions, and topic context—to build predictive models.
Linguistic Markers of Acceptance
The study found that the linguistic markers of familiarity differ significantly between entity types. For organizations, the use of indefinite articles ("a company") is a strong signal of unfamiliarity. For people, the use of appositives is a key marker of the "hearer-new" phase.
As an entity becomes more familiar, the frequency of appositives (e.g., "John Doe, a local baker") drops off. The study also identified a "definiteness paradox." While familiar entities generally use more definite articles ("the"), for people, definiteness actually decreases near the point of acceptance. This is because writers often switch to concise possessives (e.g., "Russia's president") rather than longer definite phrases ("the president of Russia").
Predicting "Household Name" Status
Using Multi-Layer Perceptron (MLP) and SVM classifiers, the researchers demonstrated that it is possible to predict an entity's future status using just a one-month "snapshot" of its current descriptions.
The MLP model achieved 74% accuracy for people and 68% for organizations in predicting when an entity would become common knowledge (within a year, 1–13 years, etc.). In a simpler binary task—predicting if an entity would be "hearer-old" within two years—the models reached approximately 80% accuracy. Syntactic features proved to be the most impactful; removing them resulted in a 5–8% drop in performance.
Conclusion and Future Work
This research provides a data-driven look at how language adapts to the growth of collective memory. While the study focused on the editorial style of The New York Times, the underlying patterns of descriptor shrinkage and syntactic shifts offer a blueprint for understanding how new concepts and figures enter our shared lexicon. Future work may expand this analysis to common nouns (like "cryptocurrency") or explore "semantic drift"—how the actual content of descriptions evolves as an entity's role in society changes.
Summary updated