We address for the first time unsupervised training for a translation task with hun dreds of thousands of vocabulary words. We scale up the expectation-maximization (EM) algorithm to learn a large translation table without any parallel text or seed lex icon. First, we solve the memory bottle neck and enforce the sparsity with a sim ple thresholding scheme for the lexicon. Second, we initialize the lexicon training with word classes, which efficiently boosts the performance. Our methods produced promising results on two large-scale unsu pervised translation tasks.
展开▼
