... input two monolingual corpora and per-haps some seed translations, and we produce as out-put a bilingual lexicon, defined as a list of word pairs deemed to be word- level translations. Preci-sion ... the word type features from a base-line normal distribution with variance σ2IdS, withhyperparameter σ2 0; unmatched target wordsare similarly generated.If two word types are truly translations, ... performanceof the system, in English-Spanish induction with ourbest feature set, using corpora derived from topicallysimilar but non-parallel sources, the system obtains89.0% precision at...