Code for the paper "Modeling Real-Time Interactive Conversations as Timed Diarized Transcripts"

Code to convert Facebook Messenger chat dumps can be found in utils/chat_utils.py. Run load_chat to load the dump and then convert_chat to compute message prefixes. See messenger/interactive_messenger.py, which implements Algorithm 1, for an example.

Code to convert WhisperX speech transcriptions can also be found in utils/chat_utils.py. Run load_transcriptions to load a directory of transcripts and then convert_transcriptions to compute prefixes. See speech/interactive_speech.py for a worked demo.