| 摘要 |
<p>The invention relates to a method of translating input data AVin into an output lexical sequence OUTSQ. During said method, sub-lexical entities and different possible combinations of said entities are identified as states ei and ej of first and second language models APMM and PHLM respectively. Said combinations are intended to be stored with an associated likelihood value Sij in a table TAB comprising memory areas. Moreover, each of said memory areas is intended to contain at least one combination of states (ei;ej) and is provided with an address equal to a value h[(ei;ej)] of a scalar function h that is applied to parameters specific to the combination (ei; ej). The invention can be used to limit the complexity of accessing information produced by a single transducer which is formed by a single Viterbi VM3 machine operating models APMM and PHLM.</p> |
