Apparatus and method for language translation using translation models based on context.

摘要

APPARATUS FOR A SERIES OF ORIGINAL TRANSLATE WORDS IN A FIRST LANGUAGE TO A SERIES OF WORDS TO WHICH results in a second language. A SERIES FOR ENTRY original words are generated at least two hypotheses of translation consisting EACH ONE OF A SERIES OF WORDS WHICH TRANSLATED. FOR EACH hypothesis which is translated, A MODEL OF CONTRAST LANGUAGE HAS AN ESTIMATE OF LIKELIHOOD OF SERIES OF WORDS IN THE ASSUMPTIONS TO ANSWER resulting. EVERY WORD TRANSLATED TO WHICH HAS A CONTEXT THAT HAS AT LEAST ONE ANOTHER WORD Assuming translation. It IDENTIFIES at least one alignment connecting each source word with at least one WORD TO TRANSLATE IN WHICH HYPOTHESIS. For each original word and HIPOTESIS TRANSLATION, A RESPONSE WORD HAS AN ESTIMATED CONDITIONAL PROBABILITY OF OCCURRENCE OF THE ORIGINAL WORD GIVEN THE WORD TO WHICH RESULTS IN hypothesize translation that is connected to the source word and given the Assuming CONTEXT translation of the word TO TRANSLATE THAT WHICH IS cONNECTED TO THE ORIGINAL WORD. TRANSLATION FOR EACH HIPOTESIS, CONTRAST RESPONSE TRANSLATION HAS A COMBINATION OF RESPONSES Constraste MODEL ASSUMPTIONS WORD FOR TRANSLATION AND WORDS TO WHICH RESULTS IN SERIES ENTRY original words. RESPONSE OF TRANSLATION hypothesis testing consists of a combination of contrast response language model to hypothesize translation and contrast response translation model to hypothesize translation. Hypothesize translation HAS THE BEST RESPONSE hypothesis testing TRANSLATION SALE.