pinakinathc
commented
4 years ago While I am aware of the lack of training data, is there any evaluation dataset? How is the acceptance criteria (F1 >= 0.75) decided?
Open raamb opened 4 years ago
pinakinathc
commented
4 years ago While I am aware of the lack of training data, is there any evaluation dataset? How is the acceptance criteria (F1 >= 0.75) decided?
Author Deborah Duong
Description Service that reads handwriting in general and in particular, the handwriting of doctors as it is in prescriptions and doctor’s notes, with its specialized notations and abbreviations.
Background: The Rejuve app requires that patient data be read in in a way that is convenient for the patient. This service will be part of a suite of services that input data to the app. Doctor’s handwriting recognition is needed so that a patient can take a picture of their prescriptions and a copy of their medical records with their phone. The service will need to understand basic handwriting as well as handwritten doctor’s notation, including commonly used abbreviations, as it appears in prescriptions and doctor’s notes, initially as it is in the United States in English. The service will be used to input patient data into a personalized medicine AI, as well as to explain to patients the definition of doctor’s notations in their data as well as some of the connotations and implications of these notations.
We know of no open dataset of doctor’s handwriting in United States medical records or prescriptions. The IAM dataset is presented for practice on handwriting, and a few medical abbreviation sites are given to represent what might be written.
Refrences
Training Dataset
Acceptance Criteria
Reward Amount 200000 tokens
Expiration Date 20 June 2020