Sunday, May 29, 2011

Amazon Mechanical Turks

I somehow stumbled onto a mention of an Amazon program called Mechanical Turks. Sometimes companies need work done that they don't have the resources or personnel for. They outsource these little jobs to Amazon and Amazon pays you to do them, then takes a 10% cut. (You can get cash in $10 increments deposited into your checking account, or Amazon gift card credits for as little as $1) The higher paying work seems to be mostly outsourced transcription. I always thought that I would like to be a transcriptionist, so I tried it out. WOW is it hard. I had no idea how many uhs, ums, ya knows, and likes, Americans use in their speech. Plus, it's really hard to listen to a sentence and then transcribe it when it's sprinkled with enough non-word placeholders to double the word count. Did you also realize that people don't actually speak in whole sentences? Every thought is strung together by, and then um, so it's like, and that is... It took me hours to do a fifteen minute phone conversation. 

Honestly, I only looked into this because it seemed like a neat way to make a little cash while sitting at home on my computer, but I would need to be pretty hard pressed for cash to do this again ($7 is not worth 4 hours work). I did, however, learn an INVALUABLE lesson about the way people speak.  I think it's going to be interesting to see how this experience effects my writing.

3 comments:

TirzahLaughs said...

Um...yeah, well like you know...um...I kinda...well sort of want to say that I felt like I might...mmmm

Sorry couldn't resist.
:) I use to do recordings of pharmaceutical drug conversations and label them for their legal files. Boring. But I got 10 bucks an hour.

BORING.

Tirz

S.K.Epperson said...

They need to perfect the Dragon software to automate the process. Surprised they really haven't yet.

Jimmie Hammel said...

The technology will get there one day I'm sure but I can understand why it's difficult. People use a lot of non-word sounds while speaking. Plus a software program would have a hard time distinguishing between different people's voices, especially if multiple people were speaking at the same time. For a transcription of a conversation, that would be critical.