FRESH_17_Topic_Modeling

Below are documentations of codes used for FRESH_17 audio diary transcript analysis regarding topic modeling.

Set up

Run the shell scripts on the cluster to preprocess the transcripts into files that can be converted to corpus in TM analysis. Run the R codes on your local computer for TM analysis (have not resolve how to run R on the cluster).

Flow of execution

1. $ bash topic_modeling_transcript_level.sh STUDYNAME SUBJECT

• calls phone_transcript_allFeatures.py if allFeatures doesn't exist
  • outputs allFeatures containing audio qc, transcript qc, and NLP summary.
• calls transcript_level_text.py
  • outputs processed daily transcript texts for the subject (Have the potential to be modified to output weekly-level texts).

2. $ bash topic_modeling_subject_level.sh STUDYNAME

• removes previously generated output file if it exists.
• calls subject_level_text.py
  • loops through and summarizes all avaliable daily transcript texts of different subjects