Not My President: How Names and Titles Frame Political Figures

Esther van den Berg, Katharina Korfhage, Josef Ruppenhofer, Michael Wiegand, Katja MarkertNAACL 2019

TL;DR

This paper investigates how the way political figures are named and titled in social media (specifically Twitter) influences public perception. They found that more formal naming conventions correlate with more positive sentiment towards the president, suggesting that naming acts as a subtle framing mechanism that emphasizes status.

What Problem Does This Paper Address?

Framing is a well-studied concept in communication and political science, focusing on how information is presented to shape audience perception. However, most research has concentrated on framing of issues and events, with less attention paid to how entities, particularly political figures, are framed. Furthermore, while sociolinguistics has long recognized names and titles as markers of status and solidarity, this has rarely been studied on a large scale or within the context of social media data. This paper bridges this gap by quantitatively analyzing how naming and titling in tweets about presidents influence sentiment, exploring the "entity framing" of political figures.

Methods

The researchers collected a corpus of tweets mentioning presidents of six G20 countries. They then developed a methodology to categorize naming forms based on formality, ranging from first name only (FN) to title and full name (TFNLN). To ensure the tweets were actually about the intended political figure and not a namesake, they employed crowd-sourced judgments. Stance annotation was also collected from multiple crowd-workers, with a robust process for identifying and removing unreliable annotators. Finally, they used statistical tests (Kruskal-Wallis and Spearman's rank-order correlation) to examine the relationship between naming formality and tweet stance.

Datasets

The primary dataset is a newly created corpus of 4002 English-language tweets mentioning presidents of France, Indonesia, Russia, South Africa, Turkey, and the United States. These tweets were collected between June and August 2017 and were annotated for stance and on-target/off-target judgments. The authors make this corpus publicly available.

Key Insights and Ideas

  • Naming as a Subtle Framing Device: The study demonstrates that even seemingly minor choices in how a political figure is named can subtly influence how they are perceived. This highlights the power of "entity framing" in shaping public opinion.
  • Formality Signals Status: The core finding is that more formal naming conventions (e.g., using titles and full names) are strongly associated with more positive sentiment in tweets. This supports the sociolinguistic theory that naming is a marker of status, where formality signals higher status and respect.
  • Social Media as a Rich Data Source: The research successfully leverages social media data (Twitter) to conduct a large-scale, quantitative analysis of sociolinguistic phenomena, offering insights that might be difficult to obtain through traditional methods.
  • Beyond Simple Sentiment: While sentiment analysis is a component, the paper goes deeper by examining how the sentiment is framed, linking it directly to the linguistic choices made in naming.

Key Results

  • A statistically significant positive correlation was found between the formality of naming and the stance of tweets towards presidents (Spearman's rank-order correlation: $r_s = .32$, $p = .001$).
  • Tweets using more formal naming forms (e.g., Title + Full Name) generally exhibited more positive stances compared to those using less formal forms (e.g., First Name only).
  • A chi-square test confirmed that the presence of a title in a naming form significantly impacted stance across all six country subcorpora.
  • The study rejected the null hypothesis that naming and stance are unrelated, providing strong evidence for the framing effect of naming.

Limitations and Future Work

The authors acknowledge several limitations:

  • Address vs. Reference: The study did not distinguish between names used for direct address (e.g., "Hey, Donald!") and those used for reference (e.g., "Donald is doing X"). This distinction can be blurred on Twitter.
  • Rare Naming Forms: Some naming forms, like "First Name only," occurred rarely, making it difficult to draw firm conclusions for those specific forms in all country subsets.
  • Time Span: The data was collected over a limited period, meaning naming trends might have been influenced by specific, contemporaneous events.
  • Language Specificity: The study was limited to English tweets, and findings might differ in other languages or cultural contexts.
  • Demographics: The study did not deeply analyze how annotator demographics might influence stance annotation.

Future work suggested includes investigating the address vs. reference distinction, exploring naming effects in other languages and for different politician subgroups (e.g., female politicians), and using the created corpus for further NLP research in stance detection.

Summary updated

BibTeX
@inproceedings{berg2019not,
  author = {Esther van den Berg and Katharina Korfhage and Josef Ruppenhofer and Michael Wiegand and Katja Markert},
  title = {{Not My President: How Names and Titles Frame Political Figures}},
  booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2019},
}

Loading related content...

Figure