Closed joshkyh closed 3 years ago
Fine narratives test cases were already implemented in stage_3/test_main.py. This ticket is paused until coarse is ready for testing.
Re test fine narratives -- with the newly added head nodding and forward learning. Please confirm the first twenty talkturn do have head nodding and forward leaning.
We need reproducible tests, therefore, we need all digital assets that can reproduce the test.
Create a new directory called test_files, and put the BYO transcript and the first twenty talkturns' video into the same folder. This is the basis of end-to-end test.
Edited the dir_config.py to add the folder for test files. Test main script, removed the scripts there, and use the BYO test transcript file.
Once this done, please proceed on to test without the BYO test script i.e., using Google Speech to Text. We aim to have a two test cases, with and without BYO transcript.
(1) BYO Test script has v, vp, vpa. You have shifted the grouth truth for the test cases into 1 csv file.
Leaning forward got stuck without the BYO Test script. (2) Without BYO test script we only test for v. Let's create another csv file, for this test case without BYO transcript.
Regarding Google Speech to Text not being versioned and hence giving a different result each time. The test for Speech to Text should be relaxed (less stringent) than the BYO transcript case.
The two test cases for fine narratives have a problem that interacts with the if statement that tests for the presence of weaved talkturns.csv. Marriane is suggesting that we have a paramter that specifies whether google STT runs, and if it does, it would store the results in another file, such that it would not affect the detection of previous work.
Test case for fine done. Not doing Coarse
This is implemented already in weave_session_transcript. Proceeding to write test cases for the fine narratives.