[cs-talks] Upcoming Oral Exam on Wednesday 27th
fgreen1 at bu.edu
Thu May 21 15:13:13 EDT 2015
Embedding Crowdworkers inside of Machine Learning Architectures
Mehrnoosh Sameki, Boston University
Wednesday, May 27, 2015 at 11:00am in MCS 148
Abstract: Crowdsourcing platforms empower individuals and businesses to gather large amounts of human input in a short amount of time. In my oral exam, I will focus on hybrid crowd-machine learning models which can contribute to two main research problems; First, while individuals can leverage crowdsourcing to rapidly and inexpensively collect annotations, they often face concerns about the quality of crowd work. In this talk, I will discuss common quality control approaches in crowdsourcing experiments, and focus on building predictive models of task performance using behavioral traces of crowd workers. Second, I will focus on the task of delineating boundaries of objects in images (segmentation), and discuss how to use the results of crowdsourcing experiments to train a machine learning system to determine automatically which types of images can be analyzed accurately and efficiently with or without human involvement.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks