Sentiment analysis of political communication: combining a dictionary approach with crowdcoding

Area of Science:

• Social science research
• Computational linguistics
• Digital humanities

Background:

• Automated sentiment analysis is crucial for understanding public opinion and political discourse.
• Existing tools are primarily English-based, limiting cross-lingual research.
• Digital text data offers vast opportunities for social science but requires effective analysis tools.

Purpose of the Study:

• To present a procedure for building fine-grained sentiment dictionaries using crowd-coding.
• To enable sentiment analysis in languages and domains beyond English.
• To facilitate the analysis of large text corpora efficiently.

Main Methods:

• Developed a crowd-coding procedure to collect sentiment scores.
• Constructed a negative sentiment dictionary for a chosen language and domain.
• Calculated sentence tonality using dictionary word sentiment.
• Validated dictionary-based sentiment estimates against manual coding.

Main Results:

• The crowdbased dictionary provides an efficient and valid method for sentiment measurement.
• Sentiment analysis can be performed on large text corpora previously unmanageable by manual coding.
• Empirical examples demonstrate the dictionary's utility in analyzing political statements and media reports.

Conclusions:

• Crowd-sourced sentiment dictionaries are effective for automated text analysis.
• This method overcomes the limitations of existing English-centric sentiment analysis tools.
• Enables innovative social science research on diverse textual data.