Programmatic access to mturk batches (and their results)

[danyaljj]

Hi there! 👋

First, thank you for the active maintenance of this library!

Question: is there a programmatic way to iterate over the past [finished] annotated batches that are visible in the graphical interface of mturk? (example screenshot below)

I checked the current API (https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/mturk.html) and it seems that the answer is a "no" (though I might be missing something).

[aBurmeseDev]

Hi @danyaljj, thanks for reaching out! Have you tried the existing list_* commands from the doc? It doesn't look like there are any specific commands available for processing batches. However I saw some users mentioned a workaround in a StackOverflow post that might help.

[danyaljj]

To my understanding, none of the list_* methods really address this question.

The StackOverFlow response suggests using list_hits(.) method which would work in theory. The issue is that this method does not return disposed HITs (those that are older than 120 days), even though one can see very old annotations in the graphical interface.

[aBurmeseDev]

In that case, I can confirm that only HITS with 4 Lifecycle "undisposed" States can be retrieved using ListHits and other list_* methods, as mentioned here and here:

At any point after a HIT is created in Amazon Mechanical Turk (Mechanical Turk) up until it is disposed, you can retrieve the status of a HIT using the GetHIT operation. In addition, you can call ListHITs, ListHITsForQualificationType, or ListReviewableHITs to retrieve HIT information. The HIT data structure returned by these operations includes the attributes and question used to create the HIT, as well as attributes that describe the current status of the HIT.