Measuring political and economic uncertainty: a supervised computational linguistic approach

In this paper, we develop a computational linguistic approach based on supervised machine learning using the People's Daily to measure Chinese official relations and political uncertainty towards the US. In the first step, we create training samples by asking experts to manually annotate news articles. In the second step, we use supervised machine learning algorithms to adjust our single neural network and support vector machine classifiers to better fit our training data. Finally, we combine our two individual classifiers and a dictionary approach to automatically detect whether an article in the newspaper sample is relevant. Using all of the relevant textual data, we then apply the computational linguistic approach to generate state-of-the-art indices and show that our indices outperform similar current textual indicators in some situations, particularly in the financial market.

The online version contains supplementary material available at 10.1007/s43546-022-00209-2.