|Sarah Ita Levitan|
This week I met Professor Hirschberg and the members of the Speech Lab, where I will be working this summer. I will be processing data consisting of recordings of Supreme Court sessions, and my first task is to attempt to excise the noisy parts - including coughing, rustling papers, and the occasional laughter. In order to do this, I will be using Amazon Mechanical Turk, posting short audio clips and asking the workers whether they contain noise. I started reading Amazon's documentation of their tools, and learned the terms and the required input. Then I began preparing the various components necessary to post a HIT (Human Intelligence Task). We initially planned to generate dynamic web pages for Mechanical Turk, but then hit upon a simpler solution, using a Python program to generate static pages. I first wrote a program to iterate through the sound files and group them into randomized sets, and next week I will work on generating an HTML page for each set.