The Value of Numbers in Clinical Text Classification

Clinical text often includes numbers of various types and formats. However, most current classification approaches do not take advantage these numbers. This study aims to demonstrate that using as features can significantly improve the performance models. also demonstrates feasibility extracting such from clinical text. Unsupervised learning was used identify patterns number usage in These were analyzed manually converted into pattern-matching rules. Information extraction incorporate a document representation model. We evaluated models trained on representation. Our experiments performed with two (vector space model word embedding model) (support vector machines neural networks). The results showed even handful numerical performance. conclude commonly representations represent way machine algorithms effectively utilize them features. Although we demonstrated traditional information be effective converting features, further community-wide research is required systematically process.