Karandeep Singh, MD, MMSc, with the University of Michigan presents on using R Package for clinical natural language processing (NLP) using sPacy. Singh covers the historical use of technologies in NLP for clinical trials being Java-centric and based on unstructured information. The need for less configurations and lack of seamless integrations with data science languages, bring in sPacy. sPacy is an industrial-strength natural language processing system used to alleviate difficulties and do more than other clinical tools. The team further presents on {clinspacy} and the goals for easier set up, simplifying clinical text, tidier results, and much more.

Originally presented at R/Medicine 2021 by Jie Cao and Karandeep Singh
https://r-medicine.org

Main Sections

0:00 Introduction
1:07 Along came spacy
2:04 sPacy - More than NLP tools
3:33 sPacy is not sufficient on its own for clinical NLP
8:31 The goals of {clinspacy}
10:40 What {clinspacy_init() does under the hood
12:06 Make annotations of clinical text simple
13:14 What if input is a data frame?
14:03 A tale of two diabetes
14:35 Let’s map all entities to UMLS codes
15:49 Entity linking is not perfect
16:07 How do we put predictors in a model?
17:11 How do we use embeddings
17:34 A 3-dimensional word embedding
18:02 Clinspansy includes 2 sets of entity embeddings
18:49 Entity embedding
19:02 How would we include the first 10 embeddings as predictors?
19:52 For more details/ Thank you

More Resources

Main Site: https://www.r-consortium.org/
News: https://www.r-consortium.org/news
Blog: https://www.r-consortium.org/news/blog
Join: https://www.r-consortium.org/about/join
Twitter:   / rconsortium  
LinkedIn:   / r-consortium