Theory-Driven Analysis of Large Corpora: Semisupervised Topic Classification of the UN Speeches

There is a growing interest in quantitative analysis of large corpora among the international relations (IR) scholars, but many of them find it difficult to perform analysis consistently with existing theoretical frameworks using unsupervised machine learning models to further develop the field. To solve this problem, we created a set of techniques that utilize a semisupervised model that allows researchers to classify documents into predefined categories efficiently. We propose a dictionary making procedure to avoid inclusion of words that are likely to confuse the model and deteriorate the its classification performance classification accuracy using a new entropy-based diagnostic tool. In our experiments, we classify sentences of the United Nations General Assembly speeches into six predefined categories using the seeded Latent Dirichlet allocation and Newsmap, which were trained with a small seed word dictionary that we created following the procedure. The result shows that, while keyword dictionary can only classify 25% of sentences, Newsmap can classify over 60% of them accurately correctly and; its accuracy exceeds 70% when contextual information is taken into consideration by kernel smoothing of topic likelihoods. We argue that once seed word dictionaries are created by the international relations community, semisupervised models would become more useful than unsupervised models for theory-driven text analysis.

Automated summary

There is a growing interest in quantitative analysis of large corpora among international relations scholars. To address the challenge of using unsupervised machine learning models consistently with existing theoretical frameworks, researchers have proposed a set of techniques that utilize a semisupervised model for efficient document classification. This approach involves creating a dictionary and using an entropy-based diagnostic tool to improve classification accuracy. Experimental results demonstrate the superiority of semisupervised models over unsupervised models, particularly when considering contextual information.