how to compute the probability of a sentence?

[fanc...@gmail.com]

I am new to language model. I want to compute the probability of a sentence using language model. Other people have trained a language model and save it as arpa format(using other tool)

I can convert it to berkeleylm's internal format using MakeLmBinaryFromArpa. So I have a NgramLanguageModel instance in java.

in the source code:

/**

* Scores a complete sentence, taking appropriate care with the start- and

* end-of-sentence symbols. This is a convenience method and will generally

* be inefficient.

* 

* @return

*/

public float scoreSentence(List<W> sentence);

it seems this method is what I want. But the doc says it's not efficient. So what should do to make it efficiently?