ANN de-identification systems require users to install deep learning frameworks, download pre-trained word vectors, and convert target text files into a particular format. These requirements necessitate substantial programming skills, may overwhelm end users, and limit the adoption of ANN systems. To overcome these barriers, we developed a ready-to-use ANN app, NDeID, to de-identify clinical text without any prerequisite installation or training. To the best of our knowledge, NDeID is the first standalone ANN de-identification app.
Describe the new knowledge and additional skills the participant will gain after attending your presentation.: 1. The existence of the first standalone ANN de-identification application, NDeID
2 Performance comparison among state-of-the-art de-identification systems and NDeID
3. Overall steps about how to build standalone application from trained ANN model
4. General limitations of standalone ANN applications
5. How to use NDeID and where to get it.
Kahyun Lee (Presenter)
George Mason University
Özlem Uzuner, George Mason University
Mehmet Kayaalp, National Library of Medicine