NXT Switchboard xml parsing script

Description:

file_list.txt is a list of file indices. It is obtained by running filelist.py script.

filelist.py read through "terminal" folder to read the file indices out.

parse_all.py This is the main script to retrieve plain text information from corresponding folders One dialogue per running. This script will take in one argument, which is the unique index of the dialogue file (e.g. sw2005)

parse_all.sh This is the bash script to run parse_all.py through all dialogue files, and list result in plain text with the index of the dialogue(e.g. sw2012)

How to run:

First, put the file_list.txt parse_all.py and parse_all.sh into .../xml/ directory

Second, mkdir /result in the .../xml/ directory

Finally, run

./parsing_all.sh file_list.txt

here file_list will guide paring_all to create corresponding result files.

Data split: The whole dataset will be divided into 3 parts ,namely training set valid set and test set, according to The NXT-format Switchboard Corpus by Sasha Cohoun etc.

This repo is under MIT licence