ACL2014 's best paper Fast and robust neural Network Joint Models to statistical machine translation uses deep learning knowledge in SMT, and presents a Neura L Network Joint Model (in fact, the fusion of source-end language knowledge to do language models), the results of the experiment presented in this paper claimed to greatly improve the previous translation performance based on N-gram language model translation system.
The Moses of the Open source machine translation system is also implemented and integrated. See [1]
An implementation of Devlin et al., a neural network language model, uses a Target-side history as well as source-side Context, is implemented in Moses as BILINGUALLM. It uses nplm as back-end (check its installation instructions).
But Moses himself named him bilingual neural LM (bilingual neural Network language model).
As with other language models, Moses treats it as a translation feature (feature) into its own decoder. The integration of features is described in [1]. But if impatient, just change the Moses configuration file Mose.ini, in the training and decoding phase, using the previous common Moses decoder (usually using KENLM or srilm as a language model), the Moses decoder will not be able to identify the error feature.
See [1], the correct procedure is:
1. Install the Neural Network Language Model tool (a Neural network language Model Toolkit) NPLM.
2. Take parameters at compile time
- --with-nplm=<root dir of the NPLM toolkit>
But I encountered a problem during the installation process:
unable to load Boost.Build:could not find" Boost-build.jam "
Attempted search from/path/to/mosesdecoder up to the root at/path/to/mosesdecoder/share/boost-build. And in these directories from Boost_build_path and boost_root:/usr/share/boost-build
At first, I've been thinking of the Boost library installation error. Refer to [2],[3] and so on, to boost a burst of toss. Without any effect ....
Reference:
[1] Http://www.statmt.org/moses/?n=FactoredTraining.BuildingLanguageModel#ntoc38
[2] http://comments.gmane.org/gmane.comp.nlp.moses.user/6322
[3] http://comments.gmane.org/gmane.comp.nlp.moses.user/11461
Copyright NOTICE: This article for Bo Master original article, without Bo Master permission not reproduced.
[Moses notes] compiling Moses decoder with NPLM