From fb1db29bdb9a09559bc7ae59ddecacb8ecff0854 Mon Sep 17 00:00:00 2001 From: agrawalm Date: Tue, 17 Nov 2020 01:25:45 -0500 Subject: [PATCH] Update index.md --- docs/index.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/index.md b/docs/index.md index f048f2c..d5ed3cd 100644 --- a/docs/index.md +++ b/docs/index.md @@ -12,12 +12,12 @@ However, this data often exists only in clinical text, and not in any structured ## Features ### Pre-Annotations Sometimes concept mentions are simple and straightforward for algorithms to recognize. In these cases, pre-annotations can save annotators time and energy. Pre-annotations outline the suggested span of text alongside its predicted label. The user can then to choose to accept the machine suggestion in just a single click, or modify or delete. In the image below, there are 3 pre-annotations: on flagyl, BUN, and NGT. - +
PRAnCER can flexibly set the pre-annotations to the outputs of any clinical entity extraction system (MetaMap, cTAKES, ClinicalBERT). A user just provides the spans and expected labels in a CSV file; we provide scripts to generate these CSV's from dictionary lookups and from scispaCy. ### Recommendations Even when a model can't settle on a single label with high-confidence, it can often surface a correct label in its top few predictions. PRAnCER comes built-in with an NLP recommendation algorithm for suggesting likely concept labels once a span of text is highlighted. Below you can see that we can correctly recommend 'vancomycin' for the term vanco. The recommendation function is merely a Python call, so one can easily swap out our recommendation algorithm for any new model. - +
### Search