Return a list of the most probable segmentations.

[rafaveguim]

It would be great if wordsegment returned that. For instance:

> ws.rank("nobodyelse")
[ ["nobody", "else"], ["no", "body", "else"], ...]

or

> ws.probabilities("nobodyelse")
[
[ ["nobody", "else"], ["no", "body", "else"], ...],
[ 0.727362, 0.0012372, ...]
]

[grantjenks]

Why? What's your use case? And how do you define "most probable"?

[rafaveguim]

Your algorithm is probabilistic, or at least it uses some sort of score. If I'm using wordsegment as part of a language modelling pipeline, I may want to propagate that measure of uncertainty.

I'm using it to model language in passwords. After a password is segmented, part-of-speech and semantic tags are inferred. The model sees these features and updates its beliefs. The devil is in ambiguity: therapistfinder has two meanings depending on segmentation. Ideally, the model should account for both.

[grantjenks]

Ok, pull request welcome.