HQSYN16

I will need a python script assigning new PWord tags to a sequence of phones (or preferably ASF units). Then I can put it into the scripting framework. Reduction of sentences to 20 words should not be a big problem.

File pros_words.py contains function set_pwords( unit_list, ver=1 ) where

• unit_list is a list of units == dicts with the same keys as in ASF files (phone, pwordBoundPos, pphrsBoundPos, prosodeme and word should be set; pwordBoundPos and prosodeme are updated)
• ver is the version of prosodic words
  • 1 .. joins monosyllabic prepositions, conjuctions a, i and enclitic pronouns se, si
  • 2 .. joins also monosyllabic relative pronouns

File asflight.py is just for testing --> pros_words.py can be used directly for the processing of ASF files.

After a week of troubles, I have finally managed to get the experiment working.
Resulting logs are stored to Experiments/pros_words/out_${spkr}_${ver} directories, where ver = 0 is for the baseline system and ver = 1,2 are for the individual versions as described in #3941#note-2.

Please, check the logs, if the position in prosodic words (namely begWeight, midWeight and endWeight attributes) and prosodemes are assigned correctly to units!

After fix of the prosodeme type handling in TC computation, there are new logs in Experiments/pros_words/out_${spkr}_${ver} directories (same as before).