Skip to content

Latest commit

 

History

History
135 lines (103 loc) · 7.64 KB

results.md

File metadata and controls

135 lines (103 loc) · 7.64 KB

Home > Analyzing Data

Analyzing Data

When your Batch is finished, download the answers either using the AMT website or the mturk_utils script (depending on the method used to create the Batch).

  1. Modify condition_pattern in your result parser config file i.e.YOUR_PROJECT_NAME_ccr_result_parser.cfg which was created in the first step (preparation).

    Note: In case there is possible to have a condition level aggregation in your dataset, uncomment the condition_pattern and condition_keys.

    Note: The condition_pattern specifies which part of the clip URL refers to the condition name/number that they are representing. Clips with the same value on that position, are considered to belong to the same condition and votes assigned to them will be aggregated to create the per_condition report. Example: Assuming D501_C03_M2_S02.wav is a file name,and 03 is the condition name. The pattern should be set to .*_c(?P<condition_num>\d{1,2})_.*.wav , and the condition_keys to condition_num.

    Note: You can activate the automatic outlier detection method per condition. To do so open YOUR_PROJECT_NAME_ccr_result_parser.cfg, in section [accept_and_use] add outlier_removal: true. The z-score outlier detection method is used i.e. all votes with an absolute z-score of 3.29 or higher will be considered as outlier and removed. It is specifically helpful in the CCR test.

  2. Run result_parser.py

    cd src
python result_parser.py ^
    --cfg your_configuration_file.cfg ^ 
    --method ccr  ^
    --answers downloaded_batch_result.csv ^
    --quantity_bonus all ^
    --quality_bonus
    • --cfg use the configuration file generated for your project in the preparation step here (i.e.YOUR_PROJECT_NAME_ccr_result_parser.cfg).
    • --method could be either acr, dcr, ccr, p835, pp835 or p804.
    • --quantity_bonus could be all, or submitted. It specify which assignments should be considered when calculating the amount of quantity bonus (everything i.e. all or just the assignments with status submitted i.e. submitted).

    Beside the console outputs, following files will be generated in the same directory as the --answers file is located in. All file names will start with the --answers file name.

    • [downloaded_batch_result]_data_cleaning_report: Data cleansing report. Each line refers to one line in answer file.
    • [downloaded_batch_result]_accept_reject_gui.csv: A report to be used for approving and rejecting assignments. One line for each assignment which has a status of "submitted".
    • [downloaded_batch_result]_votes_per_clip.csv: Aggregated result per clip, including MOS, standard deviations, and 95% Confidence Intervals.
    • [downloaded_batch_result]_votes_per_cond.csv: Aggregated result per condition.
    • [downloaded_batch_result]_votes_per_worker.csv: Long format of rating per clip, includes: HITId, workerid, file, vote and condition.
    • [downloaded_batch_result]_quantity_bonus_report.csv: List of workers who are eligible for quantity bonus with the amount of bonus (to be used with the mturk_utils.py).
    • [downloaded_batch_result]_quality_bonus_report.csv: List of workers who are eligible for quality bonus with the amount of bonus (to be used with the mturk_utils.py).
    • [downloaded_batch_result]_extending.csv: List of HITIds with number of assignment per each which are needed to reach a specific number of votes per clip.

    Note:

    • Votes in CCR test and the calculated CMOS values should be interpreted as answer to following questions: The Quality of the "processed" clip Compared to the Quality of the "reference/unprocessed" Clip is .. (Much Worse:-3 to Much Better:+3)." On the loading time of Rating Section in the HIT APP order of processed and reference clips are randomized, but the sign of vote is always corrected to answer the above-mentioned question. Order of presentation is also saved in the downloaded csv file from AMT per question in column Answer.Qx_order where x is the question's number. rp refers to refernce-processed.

    Note for P835 method:

    • for each of the Signal, Background and Overall quality scales, aggregated ratings will be stored in a separate csv file with corresponding [postfix] (i.e.sig, bak, and ovrl):

      • [downloaded_batch_result]_votes_per_clip_[postfix].csv: Aggregated result per clip, including MOS, standard deviations, and 95% Confidence Intervals.
      • [downloaded_batch_result]_votes_per_cond_[postfix].csv: Aggregated result per condition.

    • In addition a summary in the condition level will be provided for all three scales in [downloaded_batch_result]_votes_per_cond_all.

Approve/Reject submissions

Depending to how you create the HITs (using the AMT website or script) you should use the same method for approving/rejecting submission.

Approve/Reject submissions - using website.

  1. Go to “Manage”> “Results”> find your Batch and select “Review Results”.

  2. Click on "Upload CSV" and upload the [downloaded_batch_result]_accept_reject_gui.csv file.

Approve/Reject submissions - using script/API.

  1. Run the following script:

    cd src
python mturk_utils.py ^
    --cfg mturk.cfg ^
    --approve_reject [downloaded_batch_result]_accept_reject_gui.csv

Assign bonuses

  1. Run the following script with both [downloaded_batch_result]_quantity_bonus_report.csv and [downloaded_batch_result]_quality_bonus_report.csv:

    cd src
python mturk_utils.py ^
    --cfg mturk.cfg ^
    --send_bonus [downloaded_batch_result]_*_bonus_report.csv

Extending HITs

In case you want to reach the intended number of votes per clips, you may use the following procedure:

  1. During Approve/Reject submission process, select Republish rejected assignment(s) for other Workers to complete.

  2. Run the following script with [downloaded_batch_result]_extending.csv:

    cd src
python mturk_utils.py ^
   --cfg mturk.cfg ^
   --extend_hits [downloaded_batch_result]_extending.csv

    Note:

    • Extending HITs is only possible with the above API call. As a result, new assignments will be created by API call. Assignments create by API call are not visible in the website. From the report printed by script you can see how many assignments are created. In addition, you can see in your account that some amount of funds are hold for liability. However, when the assignments are finished and submitted by workers, you can review/download them in website. Until then, you may use following script call to check the status of those assignments:

      cd src
python mturk_utils.py ^
    --cfg mturk.cfg ^
    --extended_hits_status [downloaded_batch_result]_extending.csv

    • From AMT website: HITs created with fewer than 10 assignments cannot be extended to have 10 or more assignments. Attempting to add assignments in a way that brings the total number of assignments for a HIT from fewer than 10 assignments to 10 or more assignments will result in an AWS.MechanicalTurk.InvalidMaximumAssignmentsIncrease exception.