| Sarah Ita Levitan |
| HOME | ABOUT ME | MENTOR | PROJECT | JOURNAL | REPORT |
| Journal
Week 4 It works!! This week I successfully loaded the modified components to the sandbox, and to my surprise, nothing crashed! Seeing the HIT in the sandbox helped clarify a few of the details that still need improvement. For example, I realized that the workers who will be performing the tasks that we post might answer the questions at random, and then the experiment can yield incorrect results. Luckily, Amazon accounts for this possibility, and allows the requester to create a "qualification exam" that the workers must pass before being assigned a task. I listened to several audio clips and chose a few with noise and a few without, and created a qualification exam from those samples. I also noticed that some of the sound segments were really short (less than a second) and some did not contain any speech. Next week I will write a script to concatenate segments that are too short with longer segments. This is taking longer than expected, but we should being going public pretty soon! [PREVIOUS] [NEXT] |