Syllabification of last-proclitics tokens is not optimal

[ftesser]

The syllabification of some last-proclitics tokens: i.e.

Dall'epistola
Sott'occhio

are not treated properly.

<?xml version="1.0" encoding="UTF-8"?>
<maryxml xmlns="http://mary.dfki.de/2002/MaryXML" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" version="0.5" xml:lang="it">
<p>
<s>
<prosody pitch="+5%" range="+20%">
<phrase>
<mtu accent="last-proclitics" orig="Dall'epistola">
<t accent="H+L*" g2p_method="compound" merged-token="yes" ph="' d a ll e - ' p i1 - s t o - l a" pos="S" pos_full="Sfs" prosodicPositionType="nuclearNonParagraphFinal" sentenceType="decl" specialPositionType="noValue">
Dall'+epistola
<syllable accent="H+L*" ph="d a ll e" stress="1">
<ph p="d"/>
<ph p="a"/>
<ph p="ll"/>
<ph p="e"/>
</syllable>
<syllable accent="H+L*" ph="p i1" stress="1">
<ph p="p"/>
<ph p="i1"/>
</syllable>
<syllable ph="s t o">
<ph p="s"/>
<ph p="t"/>
<ph p="o"/>
</syllable>
<syllable ph="l a">
<ph p="l"/>
<ph p="a"/>
</syllable>
</t>
</mtu>
<t pos="FS" pos_full="FS">
.
</t>
<boundary breakindex="5" tone="L-L%"/>
</phrase>
</prosody>
</s>
<s>
<prosody pitch="-5%" range="-20%">
<phrase>
<mtu accent="last-proclitics" orig="Sott'occhio">
<t accent="H+L*" g2p_method="compound" merged-token="yes" ph="' ' s o tt ' O1 - kk j o" pos="S" pos_full="Sms" prosodicPositionType="nuclearParagraphFinal" sentenceType="decl" specialPositionType="endofpar">
Sott'+occhio
<syllable accent="H+L*" ph="s o tt O1" stress="1">
<ph p="s"/>
<ph p="o"/>
<ph p="tt"/>
<ph p="O1"/>
</syllable>
<syllable ph="kk j o">
<ph p="kk"/>
<ph p="j"/>
<ph p="o"/>
</syllable>
</t>
</mtu>
<t pos="FS" pos_full="FS">
.
</t>
<boundary breakindex="6" tone="L-L%"/>
</phrase>
</prosody>
</s>
</p>
</maryxml>

Others are OK, like:

<?xml version="1.0" encoding="UTF-8"?>
<maryxml xmlns="http://mary.dfki.de/2002/MaryXML" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" version="0.5" xml:lang="it">
<p>
<s>
<phrase>
<mtu accent="last-proclitics" orig="l'umanesimo">
<t accent="H+L*" g2p_method="compound" merged-token="yes" ph="l u - m a - ' n e1 - z i - m o" pos="A" pos_full="Ams" prosodicPositionType="nuclearNonParagraphFinal" sentenceType="decl" specialPositionType="noValue">
l'+umanesimo
<syllable ph="l u">
<ph p="l"/>
<ph p="u"/>
</syllable>
<syllable ph="m a">
<ph p="m"/>
<ph p="a"/>
</syllable>
<syllable accent="H+L*" ph="n e1" stress="1">
<ph p="n"/>
<ph p="e1"/>
</syllable>
<syllable ph="z i">
<ph p="z"/>
<ph p="i"/>
</syllable>
<syllable ph="m o">
<ph p="m"/>
<ph p="o"/>
</syllable>
</t>
</mtu>
<t pos="FS" pos_full="FS">
.
</t>
<boundary breakindex="5" tone="L-L%"/>
</phrase>
</s>
</p>
</maryxml>

[giuliopaci]

The situation has improved: only one accent is kept and each syllable will have a vowel (if possible) and only one. However sequence of phones containing two vowels are split in the middle without any phonetic consideration.