when applying a LM model, the default lm is evaluated as a back-off model. But how can a user tell "ngram -ppl" that the LM should be evaluated as interpolated model?
The srilm manual has cases for interpolation over multiple LMs, but how to specify interpolation for a single LM that is trained with "ngram-count -interpolate"?
Can we use This tool for punctuation modeling in turkish?
ReplyDeletethx in advance...
when applying a LM model, the default lm is evaluated as a back-off model. But how can a user tell "ngram -ppl" that the LM should be evaluated as interpolated model?
ReplyDeleteThe srilm manual has cases for interpolation over multiple LMs, but how to specify interpolation for a single LM that is trained with "ngram-count -interpolate"?
thanks
ReplyDelete