Hello from Los Angeles!

Collapse

Announcement

Collapse
No announcement yet.
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Hello from Los Angeles!

    Hello! I've been using MTurk on and off since 2014 with mixed results. I generally find that it's difficult to consistently find HITs and make it worth while and I end up turning to other means. I've never really thought to find an online community to increase my success so I'm glad a web search brought me here. I have mostly just found answers on Reddit when needed and never really got involved with contributing to the community so I look forward to sharing anything I can. As of right now I've completed 2,371 HITs with 98.5% approval and I'm having trouble finding good HITs. I have always just used the HIT Groups page with the available filters to find HITs and it often seems like there is a whole lot of nothing. I look forward to learning and hearing from everyone!

    #2
    Welcome ingthinks! I grew up on the west side of LA and my parents are still there

    You've definitely come to the right place - our folks are more than happy to share tips and tricks, etc. You've probably already seen it, but we mostly hang out on the DHT - the daily hit thread (linked above). It's perfectly find to ask questions there and even just chat.

    As for your hits/approval rating: we should be able to recommend some trusted requesters to get your ratings up to 99%. One that immediately comes to mind are the MLDataLabeler hits. Keep an eye out for their qualifications so that you can do those. Another that is proving to be quite good is Wittgenstein

    Now: to catch those hits. You need to get read up on using scripts! I'm not talking about the ones that will Amazon doesn't like: I'm only talking about the ones that will help you see what's out there in a timely way and catch them. Here's a link to our scripts page: https://www.ourhitstop.net/scripts-chart. At the very least, I would load MTS, PandaCrazy, and PandaCrazy Helper to start. The other ones are also useful, but given you don't seem to be using any, start there.

    Feel free to reach out on the DHT or in a private message to any of the moderators and welcome again

    Comment


      #3
      Wow, this changes my whole world! Thanks CrazyCatLady I didn't realize there were other tools I could use... I'm already finding much better HITs with these scripts. I guess you learn something new every day lol. I'm familiar with the MLDataLabeler HITs. I only had 500 something HITs done a few days ago and I did 1700 or so that day they posted the gender labeling tasks. That's what got me to where I am now. Unfortunately, I did about 30 of these other image labeling tasks from JANINE CHEN and they were all rejected? It totally screwed up my approval rate but once the MLDataLabeler ones went through it brought me back up to where I am now. I know for a fact that I did most of them properly, there were only a couple that were questionable when I submitted them. I'm not sure what my options are with that situation. I tried contacting the requester but have not received a response yet. Actually I can't seem to find the requester at all now. I'm just moving on for now.. I'm going to see what I can find with Wittgenstein.

      Comment


      • Fisherman
        Fisherman commented
        Editing a comment
        Take a look at our video series https://www.youtube.com/watch?v=WXngtIO5IWc&t=146s

      • CrazyCatLady
        CrazyCatLady commented
        Editing a comment
        Oh yay, I'm so glad the scripts are working! That should help you no end.

        I also got burned by Janine Chen, but not as badly. The main thing is to keep your work for new(to you) requesters within a tolerance range. Sort of like all those "how much risk are you willing to take" questions on the surveys- you want to only do work that you know will be approved, and take risks that will keep you within a 99 or 98% approval rating.

        Speaking of approval ratings, I assume you know how to check requester approval ratings on mturk. You can also check some reviews out - take them as you will since it's usually people who had a bad experience, but it's another bit of info you can use. Turkopticon and turkerview both capture reviews. Here's a link to Janine Chen's: https://turkopticon.ucsd.edu/reports?id=A272ASCHRWO9JY

      #4
      Fisherman I just subscribed to your channel, I'll check it out tonight when I sit down for YouTube time. Thank you!

      Comment

      Working...
      X