Measuring political and economic uncertainty: a supervised computational linguistic approach

Area of Science:

• Computational Linguistics
• Political Science
• Machine Learning

Background:

• Measuring political relations and uncertainty is crucial for international affairs and financial markets.
• Existing textual indicators may not fully capture the nuances of Sino-US political dynamics.

Purpose of the Study:

• To develop a novel computational linguistic approach for measuring Chinese official relations and political uncertainty towards the US.
• To create state-of-the-art indices based on supervised machine learning analysis of news articles.

Main Methods:

• Utilizing supervised machine learning algorithms, including neural networks and support vector machines.
• Manual annotation of news articles from the *People's Daily* by experts to create training data.
• Combining individual classifiers with a dictionary approach for automated relevance detection.

Main Results:

• Generated new indices for Chinese official relations and political uncertainty towards the US.
• Demonstrated that the developed indices outperform existing textual indicators in certain contexts.
• Highlighted particular effectiveness in financial market applications.

Conclusions:

• The computational linguistic approach provides a robust method for quantifying political dynamics.
• The new indices offer valuable insights for understanding Sino-US relations and their market impact.
• This methodology advances the application of machine learning in political and economic analysis.