coding-and-transcribing.md

Coding and transcribing

These are the guidelines I follow for coding and transcribing my production studies.

• How to code and transcribe
• FAQs

How to code and transcribe

What you need

For each participant, there are three important items that you need to have.

1. The production file, found with the participant's raw-data.
2. The WAV files, found in a folder with the raw-data.
3. The data sheet, a hand-written sheet of paper from the experiment

Make copies of the production and rating files from raw-data

1. Make a copy of the participant's production file and put it int the processed-data/production folder.
2. Make a copy of the participant's rating file and put it in the processed-data/rating folder.

Here is an example of what the processed data folder should look like. This is the processed data folder for the 0104-inconinput-1day-pluralmorph-6733-training experiment.

This ensure that the raw data itself is never touched. There is no need to make copies of the audio files as you will not be altering them and they are not used for data analysis.

Setup the production file

Open the COPY you just made of the production data file in the processed-data folder for the experiment.

Updated 2016-11-08

Add 10 columns: transcription, corrVerb, corrNoun, actualDet, categoryDet, notes, exp.feedback, exp.prompt.verb, exp.prompt.noun, and exp.prompt.det.


Study specific additions

• 0167-empiricalyang-9noun-hfrule-adults-fastproduction
  • add 3 additional columns: finished.verb, finished.noun, finished.det

Transcribe and code

You will fill in each of these columns with the following for each trial:

Transcribing

• transcription: what the participant actually said

Coding

• corrVerb: did they produce the correct verb? (0 for no, 1 for yes)
• corrNoun: did they produce the correct noun? (0 for no, 1 for yes)
• actualDet: what marker (determiner) did they produce? (write it exactly)
• categoryDet: what category does the marker belong to?
  • for inconinput the options are maj, min, other, and null
  • for yang the options are R, e, other, and null
• notes: write any notes you feel would help
  • for example: the child produced ka as ko; transcriptions are missing; etc
• exp.feedback : did the experimenter provide any feedback on the trial? (0 for no, 1 for yes)
  • for example: "great job!". Count any feedback no matter how small.
• exp.prompt.verb : did the experimenter prompt the participant for the verb? (0 for no, 1 for yes)
  • for example: "it starts gentif..."
• exp.prompt.noun : did the experimenter prompt the participant for the noun? (0 for no, 1 for yes)
  • for example: "it was gentif mawg"
• exp.prompt.det : did the experimenter prompt the participant for the determiner? (0 for no, 1 for yes)
  • for example: "do you want to add one of the endings you learned?"

Study specific coding

• 0167-empiricalyang-9noun-hfrule-adults-fastproduction
  • finished.verb: did the participant finish saying the verb before the beep? (0 for no, 1 for yes)
  • finished.noun: did the participant finish saying the noun before the beep? (0 for no, 1 for yes)
  • finished.det: did the participant finish saying the determiner before the beep?(0 for no, 1 for yes)

Here is an example with some of the trials filled in.


FAQs

How do I "transcribe" the sound files for the transcription column?

Listen to the WAV file for the trial in Audacity. Write what the participant said, word-for-word, in quotes. For example, "gentif mawg ka. [Did I say it right?]". If the participant goes off topic, transcribe that part in brackets [ ]. If the participant goes off topic for a really long time (this happens often with children), just write [child talking...] in the brackets, or [teacher interrupted...]. There is no need to write every single word the participant says when it is not related to the task.

What if the participant says the same word twice?

If the participant says the same word twice, such as gentif gentif mawg ka, there are two things you can do. First, check the participant's data-sheet to see how the experimenter recorded the trial. The data-sheet is considered to be the most accurate record of the trial. Second, use your best judgment during the transcription. For example, the participant may have just stuttered as in gentif, um, gentif mawg ka. Interpret whether the participant intended to say the same word twice for the trial as best you can. Make a note of this in the notes column.

What if the data sheet does not match the transcription?

The data-sheet is considered to be the most accurate record of the trial. When there is a conflict, go with what is written on the data-sheet. Make a note of the conflict in the notes column.

What if one of the three important items are missing?

Missing production file

If the production file is missing, check inside the experiment folder on the local computer. (In the data folder.) If you cannot find it anywhere, make a note on the subject tracking sheet and recommend the subject for exclusion. No further transcription or coding is necessary.

Missing WAV files

If any or all sound files are missing, check inside the experiment folder on the local computer. (In the data folder.) If you cannot find them anywhere, make a note on the production file. Use the data-sheet to fill in the transcription column.

Missing data sheet

If the data sheet is missing, make a note on the production file. Use the WAV files to fill in the transcription column.

Missing both WAV files and data sheet

If both the data sheet and the WAV files are missing, make a note on the subject tracking sheet and recommend the subject for exclusion. No further transcription or coding is necessary in this case.

How do I know what to put in prod-det-category?

Here you will put the category of the marker that the participant used. There are typically four options. For inconinput, the options are:

• maj: the majority marker (depends on language)
• min: the minority marker (depends on language)
• other: some other word (e.g. English +s, any other word)
• null: no marker used

For yang, the options are:

• R: the regular form (ka)
• e: any exceptional form (depends on language)
• other: some other word (e.g. English +s, any other word)
• null: no marker used

If you are uncertain how to code something, just ask me in person.

What if the participant started saying something but didn't finish before the beep?

The participant must be finished with the entire word for it to count in the finished.verb, finished.noun, or finished.det categories.