Environment:
OS Debian
gcc 4.8.3
GNU make 4.0
Language:
Python
Matlab
Package:
NLTK
gensim
Directory tree and file list:
because the path is hard coded in makefile and some python codes,
you have to make sure the directory tree is the same to guarantee
proper program behavior.
/ : root folder of all programs and data
/Readme : this file
/Report.pdf : HW3 report
/Holmes_Training_Data/ : training data
/testing_data.txt : testing data
/src/ : all source code
/feature_1_100_2/ : feature generated after preprocessing
/feature_1_100_2/N : 1-of-N feature
/feature_1_100_2/Vec : word2vec feature
Usage:
if NLTK package is not installed:
$ pip install nltk
$ python -m nltk.downloader all
if gensim is not installed:
$ pip install gensim
$ python ./src/preprocess.py -f ./
After preprocessing, edit main_rnn.m to setup your parameters, input
file folder and etc. Then start matlab and run:
>> main_rnn
This will do both training and testing.
For fRNN, do the similar things as RNN and use main_frnn.m instead.