Explainable Articial Intelligence applied to eventordering

In the last few years automatic systems for codifying medical documents are emerging. The latest proposals are based on multi-label classication and deep learning techniques. One step ahead is to understand the decision made by the deep learning approach and to locate whichparts of the notes are triggering the code proposed by the system.
The aim of this project is to cope with medical terms or words that are relevant to ascertain each code in an attempt to improve natural language understanding mechanisms. The location of relevant pieces of text that motivated each code, opens a room for articial understandingdevices such as the chronological ordering of events encoded.
Learning outcomes:the student will acquire background in medical text mining reinforcing the following areas:
•deep learning applied to explainable textual classification
•chronological event ordering
•relatedness and confidence metrics for artificial intelligence in text classification
Goals:the student will apply deep learning techniques in order to build a prototype able toidentify relevant words that explain the decision made by experts and furthermore learn tomimic that decision. This work can be extended and applied to chronological event ordering.Languages:this work can be carried out in English, Spanish or Basque.