Hyperpartisanship and Affective Polarization: A Natural Language Processing Approach

Prepared for presentation at the MPSA Annual Meeting, April 13th , 2023.

Karl Ho
University of Texas at Dallas

Motivation

Motivation

  • Project: trace and track the source of partisan movements in different data, including:

    • Books

    • Entertainment (e.g. movies, songs)

    • Media

    • Social Media

  • The current study begins with collecting books and authors data.

  • Apply Natural Language Processing methods to analyze text data.

 

Motivation

  • Is Taiwan polarized?

  • Is Taiwan in hyperpartisan era?

Political Polarization: Three Party Structure

Structure of the Negative Partisan Sentiments and Exclusivity: Correlation Heatmap

  • hatekmt/hatedpp:
    • Which of the following ethnic groups do you hate the most?
  • angrykmt/angrykpp:
    • Which political party ever made you feel angry?
  • threatkmt/threatdpp:
    • From point 0-10, how many points will you grade KMT/DPP as threatening to the normal development of the country?
  • negkmt/negdpp:
    • From point 0-10, how many points will you grade KMT/DPP from "no negative impression" to "very negative impression"?
  • marriedkmt/marriedpp:
    • How would you feel if you had a son or daughter who married a KMT/DPP's supporter? Very disappointed/upset to not disappointed/upset at all?

Figure 7. Structural models of Hyperpartisans, DPP and KMT

Hyperpartisan_DPP Hyperpartisan_KMT
Chi2 (df=3) 14.738 11.206
p 0.002 0.011
RMSEA 0.019 0.016
CFI 0.999 0.999

Eigenvalue Stability and Impulse-Response Plots for the Panel Vector Autoregression

What is hyperpartisanship?

Literature:

  • Communication studies 

  • Digital Journalism

    • Alt-right, Alt-left news channels

    • Negative

    • Fake news, disinformation

  • Related concepts:

    • Affective polarization

      • Elite/party vs. mass/voter

    • Sorting

What is hyperpartisanship?

Michael Kang (Boston College Law Review 2020):

  • two cohesive and hostile camps

  • voter animosity

  • ideologically polarized

  • [the other party is] “so misguided that it threatens the nation’s well-being”

  • the out-party surpasses racial prejudice

  • [one camp is] fearful of the other side

  • antagonistic

  • internally unified teams with clear, contrary positions

  • rig the rules of the game in their favor and gouge the other party in [an] outrageous fashion.

What is hyperpartisanship?

  • Exaggerated form of partisanship or partisan extremism.
  • Beyond sorting 
  • US: 
    • Breitbart news
    • The Young Turks
    • Chapo Trap House
    • Occupy Democrats

What is hyperpartisanship?

  1. Strong or superlatively partisan attachment

    • Hyperpartisans are usually strong political party supporters who identify as close or very close to one political party.

  2. Exclusivity

    • Hyperpartisans are highly exclusive of the identified rival, opposite party or camp.

  3. Hostility

    1. Hyperpartisans treat oppositive party or camp with high level of hostility with the often use of the verbs of “hate”, “despise”.

  4. Apprehension

    • Hyperpartisans consider other rival party or camp as threatening and associate it with a negative or highly negative labels

Data

  • Collection of NYT, Google Read and Amazon bestsellers

  • Scraped bestseller data (Politics and American History section) from 2011 to early 2017

  • 741 titles

  • 238 unique authors

Methods

  • Sentiment analysis

  • Topic modeling

  • Named Entity Recognition (NER)

  • Word Embeddings

  • Deep Learning Models

  • Tools

    • TextBlob

      • Subjectivity and Polarity Scores

Subjectivity

  • The subjectivity score is a measure of how subjective or objective a piece of text is.

  • It ranges from 0 to 1, with 0 being completely objective and 1 being completely subjective. The score is calculated by counting the number of words in the text that are found in a pre-defined list of subjective and objective words (Bing 2012, Poria et al 2017).

  • The more subjective words in the text, the higher the subjectivity score will be.

     

Polarity

  • The polarity score is a measure of the emotional sentiment expressed in a piece of text.

  • It ranges from -1 to 1, with -1 being very negative, 0 being neutral, and 1 being very positive.

  • The score is calculated by summing the polarity scores of individual words in the text, which are determined based on their presence in a pre-defined list of words with positive and negative sentiment.

  • The more positive words in the text, the higher the polarity score will be, and vice versa for negative words.

     

     

NYT Bestseller authors 

Politics and American History (2011-17)

NYT Bestseller authors (top 10) 

Politics and American History (2011-16)

Author(s) count
Bill O'Reilly and Martin Dugard 62
Laura Hillenbrand 46
Malala Yousafzai with Christina Lamb 27
Chris Kyle with Scott McEwen and Jim DeFelice 21
Ta-Nehisi Coates 17
Michelle Alexander 16
Bryan Stevenson 13
Mark Owen with Kevin Maurer 11
Katherine Boo 10
Marcus Luttrell with Patrick Robinson 10

NYT Bestseller authors 

Bill O'Reilly tweets (2008-2023)

NYT Bestseller authors 

Bill O'Reilly tweets (2008-2023)

NYT Bestseller authors 

Bill O'Reilly tweets (2008-2023)

NYT Bestseller authors 

Bill O'Reilly tweets (2008-2023)

NYT Bestseller authors 

Ta-Nehisi Coates tweets (2017-2023)

NYT Bestseller authors 

Ta-Nehisi Coates tweets (2017-2023)

NYT Bestseller authors 

Ta-Nehisi Coates tweets (2017-2023)

NYT Bestseller authors 

Ta-Nehisi Coates tweets (2017-2023)

Summary

  • Project is still in concept proving stage.

  • Data collection in progress
  • NLP methods are used for processing large amount of text data. 
  • Some exploratory data analyses illustrate trends of the key opinion leaders influences
  • More need be done on other construct such as:
    • Partisan Extremism
    • Hostility
    • Exclusivity 
    • Apprehension

References

  • Liu, B. 2012. "Sentiment Analysis and Opinion Mining." Synthesis Lectures on Human Language Technologies, 5(1), 1–167. https://doi.org/10.2200/S00416ED1V01Y201204HLT016

  • Poria, S., Cambria, E., Hazarika, D., Mazumder, N., & Zadeh, A. 2017. "A review of affective computing: From unimodal analysis to multimodal fusion". Information Fusion, 37, 98-125. https://doi.org/10.1016/j.inffus.2017.02.003

  • Bird, S., Klein, E., & Loper, E. 2009. Natural Language Processing with Python. O'Reilly Media, Inc.

Thank you!

Questions or suggestions will be most welcome!

MPSA 2023: Hyperpartisanship and Affective Polarization: A Natural Language Processing Approach

By Karl Ho

MPSA 2023: Hyperpartisanship and Affective Polarization: A Natural Language Processing Approach

CGOTS 2020

  • 94