Word embedding has been studied extensively in recent work. The quality of word embedding is dependent on many factors, including the size of the input corpora, model architectures, and hyper-parameter settings. In this paper, we investigate whether a larger corpus generates better word embeddings. Particularly, the word embeddings of ICD-9 codes are learned on two different datasets at an academic medical center and these embeddings are compared intrinsically and extrinsically.

Learning Objective: How corpus size influences word embedding of ICD-9 codes

Authors:

Cheng Gao (Presenter)
Vanderbilt University Medical Center

Chao Yan, Vanderbilt University
Bradley Malin, Vanderbilt University Medical Center
You Chen, Vanderbilt University Medical Center

Presentation Materials:

• Abstract/Manuscript - 1