Hope and Fear for Discriminative Training of Statistical Translation Models

97 citations recorded by the source corpus.

How researchers see this paper

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

Stochastic Structured Prediction under Bandit Feedback (2016, methodology): “The out-of-domain baseline machine translation model was trained on 1.6M parallel Europarl data and tuned with cdec ’s implementation of MIRA [5] on out-of-domain Europarl dev2006 dev set (2,000 sentences).”
Iterative parameter mixing for distributed large-margin training of structured predictors for natural language processing (2015, background): “, 2005) and translation systems with many features (Chiang, 2012) using stochastic training, in which estimates of the optimal weight vector are updated after each example, rather than use batch training systems (stochastic and batch training are reviewed in §3).”
A Stable and Effective Learning Strategy for Trainable Greedy Decoding (2018, methodology): “search has been previously used in different ways in SMT (Chiang, 2012; Gao and He, 2013; Auli and Gao, 2014; Dakwale and Monz, 2016).”
Non-Monotonic Sequential Text Generation (2019, background): “…actions are available at any given time), we inform this oracle policy with the current learned policy, encouraging it to favor actions that are preferred by the current policy, inspired by work in direct loss minimization (Hazan et al., 2010) and related techniques (Chiang, 2012; He et al., 2012).”
Learning Structured Predictors from Bandit Feedback for Interactive NLP (2016, methodology): “The in-domain MIRA model contains 133,531 active features, the out-of-domain MIRA model 214,642.”