Synchronizing transcripts/ subtitles

Tools for synchronizing transcript and subtitles. The project is developed by Emory NLP lab.

Requirement

• Install fuzzywuzzy package.

Usage

• To run the program, follow below four step.

  1. json_cleaning.py

     • extract all transcript, utterance_id, speaker from .json file.
     • input file -> .json file.
     • output file format -> pickle dumped .txt file.

  2. smi_cleaning.py

     • extract each subtitle with start/end time(milliseconds) from .smi(subtitle) file
     • input file -> .smi(subtitle) file
     • output file format -> pickle dumped .txt file

  3. smi_superset.py

     • make subtitle superset for all possible uttrance from .smi(subtitle) file.
     • input file -> pickle dumped .txt file (from step2)
     • output file format -> pickle dumped .txt file.

  4. matching.py

     • apply fuzzywuzzy matching algorithms
     • input file -> step1 output file(extracted transcript) && step3 output file(superset of subtitle)
     • output file -> result of matching

Future work