MIT algorithm designed to clear up confusion in healthcare NLP

Massachusetts Institute of Technology researchers are working on algorithms to better distinguish the meaning of words that might be used in multiple ways--a common occurrence in healthcare, reports MIT News.

Since so much data is buried in physician notes, for instance, the implications for advancing natural language processing are huge. The problem lies in "word-sense disambiguation"--determining, for example, whether  "discharge" refers to a bodily secretion or release from a hospital.

The researchers, from MIT's Computer Science and Artificial Intelligence Laboratory, will present their new system next week at the American Medical Informatics Association's annual symposium in Chicago.

The article calls the new system, with an average 75 percent accuracy rate on words with two meanings, a marked improvement over previous methods, posing the potential to make systems much more accurate and while drastically reducing the human effort required to make them.

After submitting their paper to AMIA showing a 63 percent accuracy rate, the researchers went back to the drawing board and adapted algorithms from a research area known as topic modeling. Topic modeling seeks to automatically identify the topics of documents by inferring relationships among prominently featured words. It does so by flagging clusters of words in close proximity. To that, the new algorithm adds in other aspects, such the words' syntactic roles. If an adjective appears before "discharge," for instance, it's much more likely to refer to a bodily secretion.

The algorithm also can "learn"--becoming more accurate--when unleashed on huge bodies of text without human oversight, the researchers say.

A huge thesaurus of medical terms compiled by the National Institutes of Health called the Unified Medical Language System (UMLS) are among the features the researchers plan to add into the algorithm.

Improvements in natural language processing are seen as revolutionizing medicine, offering the potential to flag potential problems found in doctors' free-text notes or to help researchers mine data linked with medical images in radiology reports.

A study published at the Journal of the American Medical Informatics Association in May highlighted the use of NLP in creating accurate clinical decision support tools.

To learn more:
- read the MIT News article


Fuel Top Line Growth and Increase Membership

In this webinar, payers can learn how to accelerate the sales pipeline and grow membership across all health insurance market segments using comprehensive sales technologies.