Clinical Decision Support through Medical Relation Extraction
Deskribapena:
This research project is devoted to Medical Relation Extraction. Examples of relations can be:
•X is the cause of Y;
•Y is located in Z;
•X was prior to W.
Medical documents contain a variety of information and, particularly, in the case of relations,the information is not always made explicit. To extract this type of information is very import when the experts have to make a decision. For example the pharmaco vigilance services must detect Adverse Drug Reactions and report the reactions to the medical community. These relations are difficult to detect. Moreover, relations can be of dierent nature and can appear in many types of documents such as: medical, legal. . . .
The aim of this project is to overcome the problem of relation extraction not as a decoupled approach in which rst the entities (X and Y in the example above) are discovered and next relations (e.g. is the cause of ) stated. By contrast, our aim is to deal with joint approaches that exploit contextual information for both entity and relation extraction. Deep neural networksoer the context to facilitate joint approaches.
Learning outcomes:the student will acquire background in information extraction reinforcing the following areas:
•deep learning applied to information extraction
•relatedness and confidence metrics for information extraction
Goals:The student will apply different techniques and relatedness metrics in order to build a prototype able to identify relations not explicitly expressed in documents.
Languages:this work can be carried out in English, Spanish or Basque.
•X is the cause of Y;
•Y is located in Z;
•X was prior to W.
Medical documents contain a variety of information and, particularly, in the case of relations,the information is not always made explicit. To extract this type of information is very import when the experts have to make a decision. For example the pharmaco vigilance services must detect Adverse Drug Reactions and report the reactions to the medical community. These relations are difficult to detect. Moreover, relations can be of dierent nature and can appear in many types of documents such as: medical, legal. . . .
The aim of this project is to overcome the problem of relation extraction not as a decoupled approach in which rst the entities (X and Y in the example above) are discovered and next relations (e.g. is the cause of ) stated. By contrast, our aim is to deal with joint approaches that exploit contextual information for both entity and relation extraction. Deep neural networksoer the context to facilitate joint approaches.
Learning outcomes:the student will acquire background in information extraction reinforcing the following areas:
•deep learning applied to information extraction
•relatedness and confidence metrics for information extraction
Goals:The student will apply different techniques and relatedness metrics in order to build a prototype able to identify relations not explicitly expressed in documents.
Languages:this work can be carried out in English, Spanish or Basque.
Non:
Bilbo
Tutorea:
Arantza Casillas & Alicia Pérez