splitted keyphrases recognized in wrong order

[fcachereul]

When splitting a long keyphrase (kewords list language model) as told in language model tutorial, the recognized hypothesis is given in reversed order. For example with the following keyphrases : oh mighty /1e-20/ powerful computer /1e-20/ If I say "oh mighty powerful computer", the recognised hypothesis is "powerful computer oh mighty"

[dhdaines]

Yes, I have also noticed this! Will try to fix for the release.