Text Analytics Showcase: Natural Language Processing and Medicine

  1 m, 44 s
We’re going to begin an occasional series showcasing emerging and interesting areas of text analysis that have caught our eye. Today’s topic: the use of natural language processing in the field of medicine.

Extracting Adverse Drug Reactions

(link was http://omop.fnih.org/sites/default/files/ omop_white_paper_friedlin_08_26_10.pdf) The FDA has an extensive digital database of medication product labels. These labels include possible adverse reactions, or negative side effects, and other important information in an unstructured format that makes access difficult, if not impossible. Two researchers used a natural language processing program to extract “ADR”s, or adverse drug reactions, from the unstructured text of the product labels. By doing so, they were able to build a database that could be used to accurately study drug and reaction associations with an accuracy upwards of 95%. Identifying Cancer (link was In this study, the authors used natural language processing on electronic medical databases to identify reports indicating cancer, as a way of replacing manual chart interpretation. They used a system that would identify clinical concepts based on the proximity of keywords within the text. Their program was able to identify the classification, location, and severity of different types of cancer, and distinguish between them and benign occurrences with a high degree of success.

Recognizing Suicidal Language in Teenagers

(link was http://www.neuron.m4u.pl/pdf/aas_2013_paper.pdf) This year, researchers at the University of Colorado School of Medicine decided to look at whether NLP machine learning could differentiate between adolescents who were suicidal, and those who were not. The algorithms they used were fed the free text responses to five different questions asked of its participants. The results? Yes, with 93% accuracy.   The myriad possibilities for role of text analysis within the medical field are staggering. NLP is helping to create medical databases, improve diagnostic accuracy, and identify illnesses, and much more. Machine learning algorithms will never replace human experts and doctors, but it may help them advance their research and save lives. We think that’s pretty cool.
Categories: Analysis, Natural Language Processing, Text Analytics