A Bilingual Graph-Based Semantic Model for Statistical Machine Translation
Authors: Rui Wang, Hai Zhao, Sabine Ploux, Bao-Liang Lu, Masao Utiyama
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The empirical results show that BGSM can enhance SMT both in performance (up to +1.3 BLEU) and efficiency in comparison against existing methods. The experiments and analysis are given in Section 5. The results in Table 5 demonstrate that BGSM is much more efficient than CSTM, especially for training, the former can be more than 50 times as fast as the later. |
| Researcher Affiliation | Collaboration | Rui Wang,1 Hai Zhao,1,2 Sabine Ploux,3 Bao-Liang Lu,1,2 and Masao Utiyama4 1Department of Computer Science and Eng. 2Key Lab of Shanghai Education Commission for Intelligent Interaction and Cognitive Eng. Shanghai Jiao Tong University, Shanghai, China 3Centre National de la Recherche Scientifique, CNRS-L2C2, France 4National Institute of Information and Communications Technology, Kyoto, Japan |
| Pseudocode | No | No pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | No | The paper mentions 'Zou et al. [2013] only released their word vectors rather than their code (http://ai.stanford.edu/ wzou/mt/)'. This refers to code from a different paper, not the authors' own source code for the methodology described. |
| Open Datasets | Yes | Corpora of IWSLT-2014 French to English (EN) [Cettolo et al., 2012], NTCIR-9 Chinese to English [Goto et al., 2011] and NISTOpen MT08 are chosen. The training data consists of part of NIST Open MT06, United Nations Parallel Text (1993-2007) and corpora of [Galley et al., 2008] that were used by [Zou et al., 2013]. |
| Dataset Splits | Yes | Corpus IWSLT NCTIR NIST training 186.8K 1.0M 2.4M dev 0.9K 2.0K 1.6K test 1.6K 2.0K 1.3K. MERT for tuning (we run MERT three times and record the average BLEU score on test data). NIST Eval 2006 is used as development data and NIST Eval 2008 as test data. |
| Hardware Specification | No | All the experiments in this paper are conducted on the same machine with 2.70GHz CPU. This provides a CPU clock speed but no specific CPU model, GPU model, or memory details. |
| Software Dependencies | No | The paper mentions 'Moses phrase-based SMT system', 'GIZA++', 'SRILM', and 'MERT' but does not provide specific version numbers for any of these software dependencies. |
| Experiment Setup | Yes | The parameters for BGSM are set as follows: 1) Vector dimensions are 6; 2) Threshold γ for edge weight pruning EW in Eq. (1) is 3 × 10−4; 3) Threshold ε for phrase table tuning DR in Eq. (5) is 1.31. The recommended settings of CSTM [Schwenk, 2012] are followed. That is, phrase length limit is set as 7, shared 320-dimension projection layer for each word (that is 2240 for 7 words), 768-dimension projection layer, 512-dimension hidden layer. The dimensions of input/output layers are the same as the size of vocabularies of source/target words. |