The Mathematics of Statistical Machine Translation: Parameter Estimation

By Peter F. Brown, Stephen Della Pietra, V. D. Pietra, Robert L. Mercer. 1993

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

Citations in the corpus

Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation (2016, background): “Neural Machine Translation (NMT) [41, 2] has recently been introduced as a promising approach with the potential of addressing many shortcomings of traditional machine translation systems.”
Transfer Learning in Natural Language Processing (2019, methodology): “…we will review unsupervised approaches which aim to model the dataset itself, brieﬂy presenting non-neural approaches (Deer-wester et al., 1990; Brown et al., 1993; Blei et al., 2003) before detailing deep neural network approaches like auto-encoding/skip-thoughts models (Dai and Le, 2015;…”
Framing Image Description as a Ranking Task: Data, Models and Evaluation Metrics (2013, methodology): “We then use Giza++ (Och & Ney, 2003) to train IBM alignment models 1–2 (Brown et al., 1993) over all pairs of noun or verb captions of the same image to obtain two sets of translation probabilities, one over nouns (P n (·|w)) and one over verbs (P v (·|w)).”
Non-Autoregressive Neural Machine Translation (2017, methodology): “These are integers for each word in the source sentence that correspond to the number of words in the target sentence that can be aligned to that source word using a hard alignment algorithm like IBM Model 2 (Brown et al., 1993).”
Recurrent Continuous Translation Models (2013, background): “We compare the perplexities of the RCTMs with the perplexity of the IBM Model 1 (Brown et al., 1993) and of the Fast-Aligner (FA-IBM 2) model that is a state-of-the-art variant of IBM Model 2 (Dyer et al., 2013).”
Statistical Phrase-Based Translation (2003, methodology): “As the ﬁrst method, we learn phrase alignments from a corpus that has been word-aligned by a training toolkit for a word-based translation model: the Giza++ [Och and Ney, 2000] toolkit for the IBM models [Brown et al., 1993].”
Modeling Coverage for Neural Machine Translation (2016, background): “Fertility To predict Φ j , we introduce the concept of fertility , which is ﬁrstly proposed in word-level SMT (Brown et al., 1993).”
Learning Semantic Correspondences with Less Supervision (2009, neutral): “This model is similar in spirit to IBM model 1 (Brown et al., 1993).”
Nearest Neighbor Machine Translation (2020, methodology): “Retrieval in Translation Recent work has integrated retrieval of words and phrases into neural translation, to gain some of the advantages of the previous generation of word-and phrase-based methods (Brown et al., 1993; Koehn et al., 2003).”
Masked Language Model Scoring (2019, methodology): “For non-end-to-end ASR and MT systems, having separate acoustic/translation models P AM / TM ( X | W ) and language models P LM ( W ) is motivated by the Bayes rule decomposition used to select the best hypothesis ˆ W (Jelinek et al., 1975; Brown et al., 1993):”
Minimum Risk Training for Neural Machine Translation (2015, background): “, word alignment, phrase segmentation, phrase reordering, and SCFG derivation) that are vital in conventional statistical machine translation (SMT) (Brown et al., 1993; Koehn et al., 2003; Chiang, 2005).”
Unsupervised Construction of Large Paraphrase Corpora: Exploiting Massively Parallel News Sources (2004, methodology): “Giza++ is a freely available implementation of IBM Models 1-5 (Brown et al. 1993) and the HMM alignment (Vogel et al. 1996), along with various improvements and modifications motivated by experimentation by Och & Ney (2000).”
Convolutional Neural Network Architectures for Matching Natural Language Sentences (2014, background): “Examples include top-k re-ranking in machine translation (e.g., comparing the meanings of a French sentence and an English sentence [5]) and dialogue (e.g., evaluating the appropriateness of a response to a given utterance[26]).”
Learning for Semantic Parsing with Statistical Machine Translation (2006, methodology): “More specifically, a statistical word alignment model (Brown et al., 1993) is used to acquire a bilingual lexicon consisting of NL substrings coupled with their translations in the target MRL. Complete MRs are then formed by combining these NL substrings and their translations under a parsing…”
Alignment by Agreement (2006, background): “Our approach is similar in spirit to co-training, where two classiﬁers, complementary by the virtue of having different views of the data, are trained jointly to encourage agreement (Blum and Mitchell, 1998; Collins and Singer, 1999).”
Modeling Interestingness with Deep Neural Networks (2014, positive): “WTM (Row 3) is our implementation of the word translation model for IR (Berger and Laf-ferty 1999; Gao et al. 2010 (Brown et al. 1993).”
Learning to Generate Pseudo-Code from Source Code Using Statistical Machine Translation (T) (2015, neutral): “Abstract syntax trees often use keywords and operators existing in the source code as internal nodes rather than leaves, but these surface words could be strongly related to speciﬁc words in the target language (e.g. the token “ if ” in Python corresponds to the word “if” in English).”