CBOW Word Embedding Example 1 context window

Automatic Detection of Vague Language in Court Decisions.


In this dissertation project, I develop a measurement for vague language in written constitutional court rulings. I use two different methods to approach this: a dictionary approach expanded using word embeddings, and a machine learning classifier. For the machine learning classifier, I use a manually annotated data set on sentences from court decisions and benchmark traditional NLP classifier versus two different types of deep learning classifiers (CNNs and RNNs). I find that while the dictionary achieves a reasonable predictive performance, the deep learning classifiers are superior in capturing semantic meaning and classifying unseen sentences.

Working Paper