# The Word2Vec Server is forked from https://github.com/3Top/word2vec-api, and did some changes to satisfy our needs.

1. Installing kaldi:
	1.1 git clone github.com/kaldi-asr/kaldi.git
	1.2 go to /tools, follow the instructions there.
	1.3 go to /src, follow the instructions there.
	1.4 go to /src/cudamatrix, change the TESTFILES to BINFILES, then make all.

2. Revise the path:
	2.1 Go to scripts/asr_run.py, revise kaldi_dir, nnet, graph to where you place kaldi, nnet_a_gpu_online and graph respectively.
	2.2 Go to scripts/run.py, revise video_dir to the DATA-files folder, txt_dir to where you want the RESULTS be, sound_dir and log_dir to the place for sound and log respectively.

2. Usage:
	Go to scripts/, 
		python3 run.py ROOT/TO/THE/DATASET.
	The result should be in the txt folder set up above, LOG files in the log folder, and the .wav files dumped from videos in the sound folder.