Difference between revisions of "User-Generated Health Content for Annotations in Watson"

From Master Projects
Jump to: navigation, search
 
Line 9: Line 9:
 
|Company=IBM
 
|Company=IBM
 
|Thesis title=User-Generated Health Content for Annotations in Watson
 
|Thesis title=User-Generated Health Content for Annotations in Watson
|Finished=No
+
|Finished=Yes
 
|Thesis=Thesisdesign.pdf
 
|Thesis=Thesisdesign.pdf
 
|Poster=thesisdesign-slides-hsk390.pdf
 
|Poster=thesisdesign-slides-hsk390.pdf

Latest revision as of 07:23, 9 December 2014


has title::User-Generated Health Content for Annotations in Watson
status: finished
Master: project within::Information Sciences
Student name: student name::Harriëtte Smook
Dates
Start start date:=2014/01/01
End end date:=2014/07/01
Supervision
Supervisor: Lora Aroyo
Second supervisor: Robert-Jan Sips
Second reader: has second reader::Chris Welty
Company: has company::IBM
Thesis: has thesis::Media:Thesisdesign.pdf
Poster: has poster::Media:thesisdesign-slides-hsk390.pdf

Signature supervisor



..................................

Abstract

IBM's Watson in healthcare serves as a clinical decision support system. Watson is currently being trained using annotations made by medical experts via the Dr. Detective game, as well as by the general crowd via micro-tasks on Amazon Mechanical Turk and CrowdFlower. These annotations come from sentences that are extracted from specialized medical texts. Nevertheless, patients often use more lay language than medical terminology, and crowdsourcing annotations via Amazon Mechanical Turk and CrowdFlower has drawbacks.

This project will investigate whether social health websites such as PatientsLikeMe, CureTogether, and MedHelp can function as alternative resources for medical data, as well as alternative platforms for crowdsourcing micro-tasks.