FAQ.md

Frequently Asked Questions

1. How to run the training code?

You can use the following command to train a model:

THEANO_FLAGS="device=gpu0,floatX=float32" python train_nea.py
	-tr data/fold_0/train.tsv
	-tu data/fold_0/dev.tsv
	-ts data/fold_0/test.tsv
	-p 1	# Prompt ID
	-o output_dir

2. How to see the available options for running train_nea.py script?

python train_nea.py -h

3. What is --emb?

You can use --emb option to initialize the lookup table layer with pre-trained embeddings:

THEANO_FLAGS="device=gpu0,floatX=float32" python train_nea.py
	-tr data/fold_0/train.tsv
	-tu data/fold_0/dev.tsv
	-ts data/fold_0/test.tsv
	-p 1	# Prompt ID
	--emb embeddings.w2v.txt
	-o output_dir

4. What is the file format of embeddings.w2v.txt?

The format of this file is the simple Word2Vec format. The first line should include the number of rows and columns of the word embeddings matrix.

5. Which pre-trained word embeddings should I use?

--emb is optinal. If you want to replicate our results, download this file. Convert En_vectors.txt to Word2Vec format and use it with --emb option. To convert it to Word2Vec format, simply add the W2V header to the file, like this:

100229 50
the -0.45485 1.0028 -1.4068 ...
, -0.4088 -0.10933 -0.099279 ...
. -0.58359 0.41348 -0.70819 ...
...