Using ranked translation choices to obtain sequences indicating meaning of multi-token expressions

摘要

To provide information about the meaning of a multi-token expression in a first language, where the information is understandable in a second language, subexpressions are obtained, such as tokens, chunks, and sentences. The multi-token expression could, for example, be a sentence or an input text with more than one sentence. Translation choices are obtained in the second language for a set of the subexpressions. A subset of the translation choices of a subexpression are ranked, and the ranked translation choices are used to produce a sequence of translation choices for the multi-token expression as a whole. Information is then presented to a user about the sequence of translation choices, indicating in the second language the meaning of the multi-token expression. Translation choices can be ranked based on syntactic or semantic context, based on translation source, based on confidence levels, based on how each was obtained, in accordance with a user-selected process, and so forth. Information about the sequence can be presented through a paper or display-based user interface. In response to user signals, additional translation choices can be presented or other modifications can be made. A user-requested change in ranking can propagate to similar subexpressions in the multi-word expression or, through rules or weighted confidence levels, to similar subexpressions in subsequent multi-word expressions.