Legal sentence boundary detection using hybrid deep learning and statistical models

The authors propose leveraging Deep learning models, including LegalBERT and CaseLawBERT, to detect sentence boundaries in English legal texts. After evaluating various deep learning models, the Convolutional Neural Network Model (CNN) is the top-performing model. 

A hybrid architecture that combines the strengths of both models, achieving a 4% improvement in the F1-score. (The F1 score is a metric that calculates the average of precision and recall. Both precision and recall have equal relative contributions to the F1 score. A perfect F1 score is 1 and the worst score is 0. This implies that a model with a perfect F1 score has made all correct predictions.)

The authors highlight the benefits of using deep learning models and hybrid architectures to achieve optimal performance in identifying sentence boundaries in legal texts.

Source: Journal of Big Data