Text as signal. A tutorial with case studies focusing on social media (Twitter)

Area of Science:

• Computational Social Science
• Psychology
• Computer Science

Background:

• Sentiment analysis and automated text coding are growing fields.
• The potential of using regularly sampled coded text as a signal is underexplored.

Purpose of the Study:

• To introduce signal processing techniques for analyzing regularly sampled coded text.
• To demonstrate the
• text as signal
• approach with social media case studies.

Main Methods:

• Finite impulse response filtering applied to emotion-coded tweets during US Election Week 2020.
• Changepoint detection for identifying shifts in emotional signals.
• Data interpolation, Fast Fourier Transform (FFT) analysis and filtering for personal value-coded tweets.
• Block bootstrapping for estimating signal variability.

Main Results:

• Visualization of filtered emotion signals and identification of key changes.
• Linking variations in personal value signals to significant real-world events.
• Demonstration of signal processing techniques for time-series text data.

Conclusions:

• Signal processing offers a novel framework for analyzing regularly sampled coded text.
• The
• text as signal
• approach enhances understanding of dynamic social phenomena.
• Readers gain foundational knowledge in applying signal processing to coded textual data.