Add files using upload-large-folder tool

ee0c4e2 verified over 1 year ago

2.52 kB

RDLM: relational dependency language model
------------------------------------------

This is a language model for the string-to-tree decoder with a dependency
grammar. It should work with any corpus with projective dependency annotation in
ConLL format, converted into the Moses format with the script
mosesdecoder/scripts/training/wrappers/conll2mosesxml.py It depends on NPLM for
neural network training and querying.

Prerequisites
-------------

Install NPLM and compile moses with it. See the instructions in the Moses documentation for details:

  http://www.statmt.org/moses/?n=FactoredTraining.BuildingLanguageModel

Training
--------

RDLM is designed for string-to-tree decoding with dependency annotation on the
target side. If you have such a system, you can train RDLM on the target side of
the same parallel corpus that is used for training the translation model.

To train the model on additional monolingual data, or test it on some held-out
test/dev data, parse and process it in the same way that the parallel corpus has
been processed. This includes tokenization, parsing, truecasing, compound
splitting etc.

RDLM is split into two neural network models, which can be trained with
`train_rdlm.py`. An example command for training follows:

  mkdir working_dir_head
  mkdir working_dir_label
  ./train_rdlm.py --nplm-home /path/to/nplm --corpus [your_training_corpus] --working-dir working_dir_head  --output-dir /path/to/output_directory --output-model rdlm_head  --mode head  --output-vocab-size 500000 --noise 100
  ./train_rdlm.py --nplm-home /path/to/nplm --corpus [your_training_corpus] --working-dir working_dir_label --output-dir /path/to/output_directory --output-model rdlm_label --mode label --output-vocab-size 75 --noise 50

for more options, run `train_rdlm.py --help`. Parameters you may want to adjust
include the vocabulary size of the label model (depending on the number of
dependency relations in the grammar), the size of the models, and the number of
training epochs.

Decoding
--------

To use RDLM during decoding, add the following line to your moses.ini config:

  [feature]
  RDLM path_head_lm=/path/to/output_directory/rdlm_head.model.nplm path_label_lm=/path/to/output_directory/rdlm_label.model.nplm context_up=2 context_left=3 context_right=0

  [weight]
  RDLM 0.1 0.1

Reference
---------

Sennrich, Rico (2015). Modelling and Optimizing on Syntactic N-Grams for Statistical Machine Translation.
  Transactions of the Association for Computational Linguistics.

	RDLM: relational dependency language model
	------------------------------------------

	This is a language model for the string-to-tree decoder with a dependency
	grammar. It should work with any corpus with projective dependency annotation in
	ConLL format, converted into the Moses format with the script
	mosesdecoder/scripts/training/wrappers/conll2mosesxml.py It depends on NPLM for
	neural network training and querying.

	Prerequisites
	-------------

	Install NPLM and compile moses with it. See the instructions in the Moses documentation for details:

	http://www.statmt.org/moses/?n=FactoredTraining.BuildingLanguageModel

	Training
	--------

	RDLM is designed for string-to-tree decoding with dependency annotation on the
	target side. If you have such a system, you can train RDLM on the target side of
	the same parallel corpus that is used for training the translation model.

	To train the model on additional monolingual data, or test it on some held-out
	test/dev data, parse and process it in the same way that the parallel corpus has
	been processed. This includes tokenization, parsing, truecasing, compound
	splitting etc.

	RDLM is split into two neural network models, which can be trained with
	`train_rdlm.py`. An example command for training follows:

	mkdir working_dir_head
	mkdir working_dir_label
	./train_rdlm.py --nplm-home /path/to/nplm --corpus [your_training_corpus] --working-dir working_dir_head --output-dir /path/to/output_directory --output-model rdlm_head --mode head --output-vocab-size 500000 --noise 100
	./train_rdlm.py --nplm-home /path/to/nplm --corpus [your_training_corpus] --working-dir working_dir_label --output-dir /path/to/output_directory --output-model rdlm_label --mode label --output-vocab-size 75 --noise 50

	for more options, run `train_rdlm.py --help`. Parameters you may want to adjust
	include the vocabulary size of the label model (depending on the number of
	dependency relations in the grammar), the size of the models, and the number of
	training epochs.

	Decoding
	--------

	To use RDLM during decoding, add the following line to your moses.ini config:

	[feature]
	RDLM path_head_lm=/path/to/output_directory/rdlm_head.model.nplm path_label_lm=/path/to/output_directory/rdlm_label.model.nplm context_up=2 context_left=3 context_right=0

	[weight]
	RDLM 0.1 0.1

	Reference
	---------

	Sennrich, Rico (2015). Modelling and Optimizing on Syntactic N-Grams for Statistical Machine Translation.
	Transactions of the Association for Computational Linguistics.