Towards Spoken Medical Prescription Understanding

Abstract : Prescription Management Systems (PMS) have appeared in health institutions to reduce medication errors which affect several million people worldwide each year. However, practitioners must enter information manually into PMS which decreases the time devoted to care. In this paper, we propose to provide a Natural Language interface to the PMS so that practitioners can record their prescriptions orally through mobile devices at the point of care. We briefly describe the overall approach and focus on the Natural Language Understanding process which was approached through slot-filling. To deal with the paucity of data and the imbalanced class problem, we present a method to artificially generate medical prescriptions. Experiments on the artificial and a realistic dataset with several state-of-the-art NLU systems show that the method makes it possible to learn competitive NLU models and opens the way to experiments on speech corpora.
Complete list of metadatas

Cited literature [41 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-02317503
Contributor : Ali Can Kocabiyikoglu <>
Submitted on : Monday, October 21, 2019 - 9:53:26 AM
Last modification on : Tuesday, November 19, 2019 - 8:18:30 PM

File

Sped2019___CameraReady.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02317503, version 1

Collections

LIG | CNRS | UGA

Citation

Ali Can Kocabiyikoglu, François Portet, Hervé Blanchon, Jean-Marc Babouchkine. Towards Spoken Medical Prescription Understanding. 10th Conference on Speech Technology and Human-Computer Dialogue, Oct 2019, Timişoara, Romania. ⟨hal-02317503⟩

Share

Metrics

Record views

40

Files downloads

63