=Paper=
{{Paper
|id=Vol-3878/37_main_long
|storemode=property
|title=ItGraSyll: A Computational Analysis of Graphical Syllabification and Stress Assignment in Italian
|pdfUrl=https://ceur-ws.org/Vol-3878/37_main_long.pdf
|volume=Vol-3878
|authors=Liviu Dinu,Ioan-Bogdan Iordache,Simona Georgescu,Alina Maria Cristea,Bianca Guita
|dblpUrl=https://dblp.org/rec/conf/clic-it/DinuIGCG24
}}
==ItGraSyll: A Computational Analysis of Graphical Syllabification and Stress Assignment in Italian==
https://ceur-ws.org/Vol-3878/37_main_long.pdf
ItGraSyll: A Computational Analysis of Graphical
Syllabification and Stress Assignment in Italian
Liviu P. Dinu1,3,* , Bogdan Iordache1,3 , Bianca Guita3 , Simona Georgescu2,3 and Alina Cristea3
1
University of Bucharest, Faculty of Mathematics and Computer Science, Romania
2
University of Bucharest, Faculty of Foreign Languages and Literatures, Romania
3
Human Language Technologies Research Center, Bucharest, Romania
Abstract
In this paper we build a dataset of Italian graphical syllables (called ItGraSyll). We perform quantitative and qualitative
analyses on the syllabification and stress assignment in Italian. We propose a machine learning model, based on deep-learning
techniques, for automatically inferring syllabification and stress assignment. For stress prediction we report 94.45% word-level
accuracy, and for syllabification we report 98.41% word-level accuracy and 99.82% hyphen-level accuracy.
Keywords
syllabification, stress assignment, Italian,
1. Introduction “prosodic revolution” [10] from Latin to the Romance lan-
guages – including syncope (the loss of an intermediate
Word syllabification and syllable analysis are two related syllable) and apocope (the loss of the final syllable) at a
issues of great importance in the study of language (writ- large scale – has led to major changes, but their weight is
ten or spoken). These topics have attracted a large cat- different from one idiom to another: while the Western
egory of researchers, from pure linguists, in phonetics, Romance languages manifest highly evident differences
to psycholinguists, computer scientists, speech thera- from the Latin phonological and prosodic system, and the
pists, etc. Thus, the syllable plays an important role in Eastern languages are considered to be most conservative
language learning and acquisition, speech recognition, from this point of view, Italian seems to be in between
speech production [1, 2], language similarity [3], in text [10]. On the other hand, in Latin, the relation between
comprehensibility (Kincaid-Flesch formula [4]), in speech stress and quantity grew stronger, thus short stressed
therapy, in poetry analysis [5, 6], etc. Each language has vowels progressively gained length. It is noteworthy that
its own way of grouping sounds into syllables and its own this situation is best preserved in Italian, and not in the
rules for dividing words into syllables. Linguistically, the Eastern Romance idioms: thus, in Italian stress cannot
syllable represents "the smallest phonetic trance likely skip a heavy penultimate syllable, and stress cannot fall
to receive an accent and only one" [7], and the syllabic further back than the antepenultimate syllable, a twofold
cut is seen by De Saussure [8] on the border between the characteristic feature of the Latin prosodic system. This
implosion and the explosion of the spoken sound: "If in is why we are taking Italian as a starting point for a larger-
a chain of sounds one goes from implosion to explosion, scale study, oriented towards all Romance languages. The
one obtains a particular effect which is the indication of main difference between Latin and its modern descen-
the boundary of the syllable". dants is that Latin stress was quantity- sensitive, leading
The analysis of the words’ syllabic structure also plays thus to the following rule: in polysyllabic words, stress
an important part in historical linguistics [9], not only fell on a heavy penultimate (meaning, containing a long
in diachronic phonetics and phonology, but also in lexi- vowel), otherwise on the antepenultimate. Due to the
cology. Romance comparative linguistics, in particular, collapse of vowel quantity as a distinctive feature in the
still needs a detailed overview of this aspect, as syllable, vocalic system, no Romance language has retained the
segmentation and prosody can give strong account on Latin stress rule as such [10]. As, from a statistic point of
phonetic changes that haven’t been explained yet. The view, the greatest part of the Romance lexicon is repre-
sented by penultimate stressed words, a basic automatic
CLiC-it 2024: Tenth Italian Conference on Computational Linguistics, mechanism would assign penultimate stress by default,
Dec 04 — 06, 2024, Pisa, Italy
*
Corresponding author.
whereas for both final and antepenultimate stress, the
$ ldinu@fmi.unibuc.ro (L. P. Dinu); machine (as well as, not in a few cases, non-native speak-
iordache.bogdan1998@gmail.com (B. Iordache); ers) would need further specification. As a consequence
bianca.guita@s.unibuc.ro (B. Guita); of the loss of Latin vowel quantity, Romance stress has
simona.georgescu@lls.unibuc.ro (S. Georgescu); ceased to be completely predictable. That is, partially,
alinaciobanu20@gmail.com (A. Cristea)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License why in the majority of the traditional Romance compara-
Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
tive or historical grammars, there is no specific section other linguistic factors that those rules take into account.
devoted to syllabification [11], or, if there is, it focuses For example, a rule that is present in many languages
either on general prosodic features [12], or on the vowel distinguishes between a vowel and a semivowel, but the
evolution depending on its presence in an open or closed computer is not able to easily recognize when the same
syllable [13]. The lack of a section dedicated to syllab- sign has the value of a vowel and when it is a semivowel.
ification is also common in the historical grammars of Because of this, rule-based adaptations of syllabification
Italian [14, 11, 15]. We will focus in this research only systems [26] generally have higher errors, and many lan-
on written form of words, so we will investigate only guages do not have an automatic syllabification system
the graphical syllabification and stress. By focusing on yet (for example, in the Python library, only a few lan-
the graphical syllabification and stress in Italian, we aim guages have syllabification). The last few decades have
to take a step forward towards the complete evaluation brought the first data-driven syllabification systems.
of the prosodic changes that took place in the transition However, in order to build such a system, training
from Latin to the Romance languages, and their influence data is needed, and there are many cases in which the
on the Romance phonetics and phonology. A machine- available data do not cover the whole language, and thus
learning model, capable of automatically inferring graph- the systems have different results when the test corpus
ical syllabification and stress assignment, along with the is changed.
purpose of creating a data-base containing the quanti- Starting with these remarks, our main contributions
tative and qualitative description of syllabification and are:
stress in the Romance languages, could be the first im-
portant task in the greater challenge of tracing the simi- • We propose ItGraSyll (Italian graphical syllables),
larities and differences between the Romance languages a dataset of 114, 503 Italian words, in ortho-
and, more important, between Romance and Latin. From graphic form, containing annotations for their or-
a typological point of view, the study of syllabification thographic syllabification and stress placement1
and stress can shed a new light on the universal features • We perform quantitative and qualitative analyses
that, by defining our phonoarticulatory and phonoacous- of the previously built dataset.
tic apparatus, have guided the languages’ development • We analyze stress placement in the context of the
and change. Given the promising results of this analysis, Italian syllables.
the present study can establish the basis of a research of • We propose an automatic system of syllabification
the syllable in other languages, either linguistically or for Italian words.
typologically related to Italian.
One of the studies that address automatic syllabifi-
cation in Italian belongs to Bigi and Petrone [16], who 2. Quantitative Analysis
proposed a tool that performs rule-based automatic seg-
mentation. Adsett and Marchand [17] and Adsett et al. In this section we perform various measurements regard-
[18] investigated whether data-driven approaches out- ing the syllables and stress placement of Italian written
perform rule-based approaches for a language with a words and analyze the results. We perform, on Italian,
low syllabic complexity, such as Italian. The authors an investigation similar to a previous investigations con-
reached the conclusion that even in this case data-driven ducted on Romanian by Dinu and Dinu [27], Dinu and
systems are the more appropriate approach. In terms of Dinu [28].
machine learning, the tasks of automatically inferring syl-
lable boundaries and predicting stress assignment can be 2.1. Data
naturally framed as sequence labeling problems. While
We build a dataset of Italian words starting from the
automatic syllabification has received more attention re-
online version of Dizionario italiano De Mauro,2 which
cently [19, 20, 21, 22, 23, 24], stress placement has not
provides information regarding graphical syllabification
been investigated as much [25].
and stress placement for the Italian vocabulary. Stressed
Given the complexity of syllable applications and word
syllables are also shown by having accents on the domi-
syllabification, the presence of electronic resources dedi-
nant vowel. Going further, this dataset will be referred
cated to them becomes a necessity. While native speakers
to as ItGraSyll.
of a language generally do not have great difficulty in
We performed several pre-processing steps. We
spelling words, the same cannot be said of those who
cleaned the resulted dataset by removing duplicates, pre-
learn a foreign language who often tend to apply their
fixes and suffixes in order to remain with the base word;
own rules to foreign words, and problems arise in au-
tomatic syllabification. This is because the rules of syl- 1
The dataset is available for research purposes upon request at:
labification are linguistic rules, and they cannot always https://nlp.unibuc.ro/resources.html#itgrasyll
be easily modeled by the computer when there are no 2
https://dizionario.internazionale.it/
abbreviations and unwanted punctuation marks such Index Syllable Frequency
as dots, commas, apostrophes and dashes were also ex- 1 to 23943
cluded so we can correctly process each word and its 2 re 18199
syllable division. Finally, the dataset consists of 114, 503 3 ta 12796
words in orthographic form having between one and 4 te 10987
eleven syllables. The distribution of words per number 5 si 10026
of syllables is represented in Table 1. 6 a 9142
7 co 8874
#syll. #words Examples
8 ri 8868
9 ca 8478
1 722 ai 10 ra 8388
2 5,960 àc-cia 11 na 8367
3 23,286 àb-ba-co
12 ti 8184
4 41,253 a-ba-chì-sta
13 ne 8112
5 28,357 a-bi-tà-co-lo
6 10,829 ac-cu-mu-la-zió-ne 14 men 7841
7 3,294 au-ten-ti-fi-ca-zió-ne 15 la 7175
8 650 a-e-ro-mo-del-lì-sti-co 16 di 6663
9 132 bi-o-me-te-o-ro-lo-gì-a 17 le 6555
10 16 in-tel-let-tu-a-li-sti-ca-mén-te 18 li 6176
11 5 ge-ne-ra-ti-vo-tra-sfor-ma-zio-nà-le 19 no 5748
20 lo 5479
Table 1
Number of words per number of syllables. Table 2
Top 20 most frequent syllables.
2.2. Syllables
the most frequent consonant-vowel structures are the
We identified #𝑇 𝑦𝑝𝑒𝑠𝑦𝑙 = 3730 (type syllables) in following: a) for the type syllables: cvc (25%), ccvc (20.9%),
Italian. The total number of syllables (token syllables) cvvc (7.79%). b) for the token syllables: cv (58%), cvc (15%),
is #𝑇 𝑜𝑘𝑒𝑛𝑠𝑦𝑙 = 483, 931. So, the average length ccv (7%), cvv (4.74%) and v (4.32%). Moreover, we observe
of a word measured in syllables is 𝑊 𝑜𝑟𝑑𝑠𝑎𝑣−𝑠𝑦𝑙 = that the cv structure corresponds to 40 out of the most
483,931/114,503 = 4.226. The 114,503 words are formed of frequent 50 syllables from the dataset.
#𝐿𝑒𝑡𝑡𝑒𝑟𝑠 = 1,133,515 letters (graphemes). So, the aver-
age length of a word measured in letters is 𝑊 𝑜𝑟𝑑𝑎𝑣−𝑙𝑒𝑡
2.4. Stress Placement
= 1,133,515/114,503 = 9.899.
In order to characterize the average length of a syllable We identified a total of 2,883 stressed syllables (type syl-
measured in letters, we investigated two cases: a) the lables). So, 847 syllables are never stressed. The most
average length of the token syllables measured in letters frequent 20 stressed syllables are represented in Table 3.
is: 𝐿𝑆𝑦𝑙𝑡𝑜𝑘𝑒𝑛 = 1,133,515/483,931 = 2.342 b) the type We observe that the most frequent stressed syllable (men)
syllables are formed of #𝑇 𝑦𝑝𝑒𝑆𝑦𝑙𝑙𝑒𝑡 = 13,576 letters. has a very high stress ratio (90%) when we compare the
Thus, the average length of a type syllable measured in stressed occurrences with all its occurrences (stressed
letters is 𝐿𝑆𝑦𝑙𝑡𝑦𝑝𝑒 = 13,576/3,730 = 3.639. and unstressed) in our database. While in the top 20 of
These statistics are computed for the words extracted all syllables, men is the only syllable of length 3 (on the
from the dictionary, which were considered to be equally 14th position), for stressed syllables there are a couple
weighted. This excludes any information relating to the of other syllables with a length greater than 2 (zio on
frequency of the words with respect to writing or speech. position 6 with 34% stress ratio, gia on position 19 with
For future research, large corpora of Italian texts can be 65% stress ratio).
leveraged in order to recompute these values and include We investigate stress placement with regard to syllable
frequency-based weights. structure and we provide in Table 4 the percentages of
A list of the most frequent 20 syllables is included in words having the stress placed on different positions (for
Table 2. top 5), counting syllables from the beginning and from
the end of the words as well. We observe that in most
2.3. Syllable Structure cases the stress is placed on the second to last syllable.
We identified a total of 67 different consonant-vowel
structures. The most frequent 7 structures cover almost
97% of the total. Depending on the type-token ratio,
Index Syllable Frequency Stress ratio (%) 100 cover 74% and the most frequent 150 syllables (i.e.
1 men 7120 90 4% of #𝑇 𝑦𝑝𝑒𝑠𝑦𝑙 ) cover 80% of #𝑇 𝑜𝑘𝑒𝑛𝑠𝑦𝑙 . Over this
2 ta 5809 45 number, the percentage of coverage rises slowly. 2,281
3 na 3348 40 (61%) syllables of type syllables occur less then 10 times,
4 to 3254 15 and 1,174 syllables occur only once (hapax legomena).
5 la 2978 41
6 zio 2916 76 2.5.2. Stressed Syllables
7 ti 2820 34
8 ca 2461 29 A similar trend can be observed also for the stressed syl-
9 ra 2297 27 lables. Further, we notice that the most frequent syllables
10 li 2239 36 cover a wide ratio of the total syllable frequency. For
11 ri 2100 24 example, the 10 most frequent stressed syllable represent
12 tu 2024 62
31% of the total of stressed syllables, the top 50 syllables,
13 za 2022 42
60% and the top 200 syllables, 81% of the token syllables.
14 ni 1734 40
15 tri 1458 60 The values are plotted in Figure 1, for all syllables and
16 ma 1209 25 for stressed syllables.
17 si 1144 11
18 da 1109 43 Type
0.8 all syllables
19 gia 1081 65 stressed syllables
20 mi 1052 25 0.7
Table 3 0.6
Coverage
Top 20 most frequent stressed syllables. The stress ratio indi- 0.5
cates how often out of all the occurrences of the syllable in
0.4
the corpus it appears as stressed.
0.3
25 50 75 100 125 150 175 200
Syllable %words Syllable %words Number of syllables
1st 8,611 1st 3,330
2nd 25,544 2nd 94,225 Figure 1: The coverage of most frequent syllables.
3rd 40,568 3rd 16,113
4th 25,593 4th 14
5th 9,243 5th 1 This results proves that the law is true for Italian too,
(a) counting syllables from (b) counting syllables from
a very small number of syllables cover a large part from
the beginning of the the end of the word Italian language (there are necessary only 150 syllables
word to cover 80% from language).
Table 4 3. Minimum Effort Laws
Stress placement for Italian.
In this section we discuss two minimum effort laws that
have been previously investigated for other languages
2.5. Syllables’ Usage and verify whether they apply for Italian as well.
The syllables have a less intuitive behaviour, usually a
small number of syllables cover a large part from a lan- 3.1. Chebanow
guage. This is valuable for a large category of natural Denoting by 𝐹 (𝑛) the∑︀frequency∑︀ of a word having n
languages, including English, Dutch, Romanian [28], Ko- syllables and by 𝑖 = 𝑛𝐹 (𝑛)/ 𝐹 (𝑛) the average
rean, Chinese, etc. We investigate here if this empirical length (measured in syllables) of the words, Chebanow
law is also applicable to Italian. We made this investiga- [29] proposed the following law between the average 𝑖
tion both on stressed and general syllables. and the probability of occurrences 𝑃 (𝑛) of the words
having n syllables:
2.5.1. General Syllables
(𝑖 − 1)𝑛−1
The most frequent 30 Italian syllables (when stress place- 𝑃 (𝑛) = * 𝑒1−𝑖 (1)
(𝑛 − 1)!
ment is disregarded) cover almost 50% of #𝑇 𝑜𝑘𝑒𝑛𝑠𝑦𝑙 , the
most frequent 50 syllables cover 61%, the most frequent For Italian, 𝑖 = 4.226.
(a) The probability distribution of the (b) Theoretical representation of the prob- (c) Menzerath’s Law: The more syllables in
length of words. ability distribution of the length of a word, the smaller its syllables.
words.
Figure 2: Minimum effort laws.
Model Hyphen Acc. Hyphen F1 Word Acc.
GRU for syllabification w/o stress markers 99.74% 99.69% 97.61%
GRU for syllabification w/ stress markers 99.82% 99.79% 98.41%
GRU for stress prediction — — 94.45%
Table 5
Performance metrics computed for the automatic syllabification and stress prediction on the test set. We computed accuracy
and F1 scores on the sequence labelling predictions for syllabification, in order to assess how well the model predicts the
positions where the syllables split. Word level metrics were computed for both syllabification and stress prediction; this kind
of metrics are more strict since any misplaced hyphen in the syllabification makes the entire prediction wrong.
In Figures 2a and 2b we plot the probability distribution 4. Automatic Syllabification and
of the length of words (in syllables) – the practical and
theoretical representations.
Stress Assignment
We observe that the two curves have comparable We further investigate how a deep-learning model can au-
shapes, with a more prominent peak for the probabil- tomatically infer the syllabification and stress assignment
ity distribution in Figure 2a; this peak can be influenced of Italian words, given their orthographic representation.
by the fact that it is determined based on all the words in
the dictionary, where many 4-syllable words are present.
4.1. Methodology
3.2. Menzerath Both tasks can be defined in terms of a sequence la-
belling problem, strategy which was previously success-
Menzerath’s law – later generalized by the Menzerath- ful used for Romanian[31, 32]. Let us consider, for ex-
Altmann law [30] – states that the bigger the number of ample, the word medaglione (the Italian translation of
syllables in a word, the lesser the number of phonemes the word "locket"). For syllabification we can label each
composing these syllables. In other words, Menzerath’s letter from the word either with the label 1, denoting
law expresses a negative correlation between the length that a syllable starts from that letter, or with the label
of a word in syllables and the lengths in phonemes of its 0, meaning the respective letter is not the first letter in
constitutive syllables. In cognitive economy terms, this its syllable. Similarly, for identifying the stressed vowel,
means that the more complex a linguistic construct, the we can label its position with a 1 and all other letters
smaller its constituents. The law is expressed as follows: are assigned the label 0. We thus obtain for our exam-
ple the sequence 1010100010 for syllabification and the
𝑦 = 𝛼𝑥𝛽 𝑒−𝛾𝑥 (2)
sequence 0000000100 for stress prediction (i.e. me-da-
where 𝑦 is the syllable length (the size of the constituent), gliò-ne, the o vowel is stressed).
𝑥 is the number of syllables per word (the size of the lin- With these definitions, we can now construct machine
guistic construct), and 𝛼, 𝛽, 𝛾 are empirical parameters. learning models for labelling the character sequences.
Figure 2c shows that the law is satisfied for Italian. The model we propose is a recurrent neural network
based on Gated Recurrent Units (GRU) [33]. The model ar-
chitecture is comprised from the following components:
• a character embedding layer, producing 64- 4.2. Results Anaysis
dimensional vectors for each unique character
Table 5 contains the metrics computed on the test set,
• a stacked bidirectional GRU, with 3 layers and a
using the models trained for syllabification (both with
128-dimensional hidden state; a 0.2-rate dropout
and without stress markers) and the model trained for
applied after each of the first two layers
predicting the stressed vowel. We obtained a remarkable
• 0.5-rate dropout, after the last GRU layer, along hyphen accuracy of 99.74% for syllabification without
with one-dimensional batch normalization the stress markers, and, when we add the stress markers,
• a time-distributed fully-connected layer with 256 we obtained an increasing accuracy, obtaining 99.82%.
output nodes and ReLU activation Including the stress markers into the data used for syl-
• a linear layer that projects the 256-dimensional labification improved the metrics across the board, most
vector into a single number, on which sigmoid notably with a ∼ 1% increase in word-level accuracy,
activation is applied to infer the binary labels. which considering the large amount of data, and the high
accuracy scores is a significant improvement (460 fewer
For training the models for both tasks, the dataset of
syllabification mistakes as opposed to the approach that
words is split into 50% training examples and 50% test
excludes stress markers). Regarding the stress prediction,
examples, unseen during training.
we obtained an accuracy of 94.45%. Table 6 showcases a
The loss function computed for the prediction made
series of wrong predictions generated by the models on
for a word, regardless of the task on which the model
the tests sets for stress assignment and syllabification.
is trained, is the average of two terms: the first one is
We also look into the accuracy scores computed for
the average character-wise binary cross-entropy, while
the test set, when it is bucketed based on the real number
the second one is the root mean squared error computed
of syllables of the test words. These results are shown
between the vector of predicted labels and the ground-
in Figure 3 and Table 7. For stress assignment, accu-
truth vector. The model is optimized using the Adam
racy decreases to a global minimum for disyllabic words,
optimizer [34], with a learning rate of 0.0003, no weight
then starts to increase again with the number of syllables.
decay, bath size of 32, and a LR scheduler that halves it
For the syllabification task, including the stress markers
every 5 epochs. The models are trained for 10-15 epochs.
seems to outperform excluding them in most scenarios,
For the task of automatic syllabification, we wanted
while both accuracies achieve a peak around the 5 sylla-
to check if the presence of the stress markers affects the
bles mark. This result seems to align with the distribution
performance of the model. Because of that, we trained
of syllables in the dataset, i.e. obtaining higher scores
two models: the first one was trained using the spelling
for the number of syllables with more examples. For
of the words with the stress markers removed, while the
stress assignment errors, we also investigate the place-
second one was trained with them included.
ment of the predicted stressed syllable in relation with
the true one (see Table 8). 95.6% of the errors misplaced
Stress Assignment Errors
the stressed syllable at most one position to the left, or
True Predicted to the right, while almost two thirds of the erroneous
bàlano balanò predictions placed the stress on the first syllable to the
fèmore femòre right of the correct one.
dòlmen dolmèn
tùtolo tutòlo
pudìco pùdico
corsìa còrsia 100.0
97.5
Syllabification Errors
95.0
True Predicted
Accuracy
mu-o-ne muo-ne 92.5
bion-da bi-on-da 90.0
cli-en-te clien-te
co-di-a-to co-dia-to 87.5 Task
Stress Assignment
ma-nu-brio ma-nu-bri-o 85.0 Syllabification (w/o stress markers)
spa-tria-to spa-tri-a-to Syllabification (w/ stress markers)
1 2 3 4 5 6 7 8 9 10 11
Table 6 Num. Syllables
Examples of erroneous test predictions provided by the deep- Figure 3: The test accuracies for each of the three tasks,
learning models. computed independently on the test words, bucketed by their
true number of syllables.
Num. Syllables Num. Words Stress Assignment Syllabification (w/o SM) Syllabification (w/ SM)
1 721 99.03% 83.63% 84.88%
2 5,960 92.94% 96.56% 97.80%
3 23,286 94.46% 98.55% 99.19%
4 41,253 97.42% 99.03% 99.48%
5 28,357 98.92% 99.33% 99.49%
6 10,829 99.48% 99.23% 99.26%
7 3,294 99.67% 99.15% 99.15%
8 650 100.0% 99.23% 98.46%
9 132 100.0% 99.24% 99.24%
10 16 100.0% 93.75% 93.75%
11 5 100.0% 100.0% 100.0%
Table 7
Similar to Figure 3 this table contains the actual values of the test accuracies for the three tasks: stress assignment, and
syllabification with/without stress markers (SM) included. These scores are computed separately for words with the same
number of syllables.
Stressed Syllable Delta Num. Errors Pct. Errors
-2 21 0.74%
-1 804 28.38%
0 95 3.35%
1 1,809 63.85%
2 102 3.60%
3 2 0.07%
Table 8
Starting from the incorrect predictions for stress assignment, we compute how far the assigned stress is from the actual one,
in numbers of syllables (delta). A delta of −2 means that the predicted stressed syllable is the second one to the left of the
correct stressed syllable. A delta of 0 in this situation means that the algorithm predicted the stressed vowel incorrectly, but
the prediction sits inside the correct stressed syllable.
5. Conclusions Innovation and Digitization, CNCS/CCCDI UEFISCDI,
SiRoLa project, number PN-IV-P1-PCE-2023-1701, Roma-
In this paper we have investigated graphical syllabifica- nia.
tion and graphical stress assignment for Italian words.
We have started by building ItGraSyll, a dataset of Italian
graphical syllabified words, with stress annotations as References
well, on which we have performed several quantitative
and qualitative analyses, including the verification of [1] S. Suyanto, Incorporating syllabification points into
two minimum effort laws for the case of Italian. Finally, a model of grapheme-to-phoneme conversion, In-
we have proposed a recurrent neural network machine ternational Journal of Speech Technology 22 (2019)
learning model for automatic syllabification and stress 459–470.
assignment for Italian written words. For stress predic- [2] V. N. Vitale, F. Cutugno, A. Origlia, G. Coro,
tion we have obtained 94.45% word-level accuracy, and Exploring emergent syllables in end-to-
for syllabification we have obtained 98.41% word-level end automatic speech recognizers through
accuracy and 99.82% hyphen-level accuracy. In future model explainability technique, Neural
work we intend to extend the analysis from dictionary Comput. Appl. 36 (2024) 6875–6901. URL:
level to corpus level and to investigate other languages https://doi.org/10.1007/s00521-024-09435-1.
as well. doi:10.1007/S00521-024-09435-1.
[3] A. Dinu, L. P. Dinu, On the syllabic similari-
ties of romance languages, in: A. F. Gelbukh
Acknowledgments (Ed.), Computational Linguistics and Intelligent
Text Processing, 6th International Conference, CI-
We want to thank the reviewers for their useful sugges- CLing 2005, Mexico City, Mexico, February 13-19,
tions. Research supported by the Ministry of Research, 2005, Proceedings, volume 3406 of Lecture Notes
in Computer Science, Springer, 2005, pp. 785–788. 463. URL: https://doi.org/10.1016/j.csl.2009.02.004.
URL: https://doi.org/10.1007/978-3-540-30586-6_88. doi:10.1016/j.csl.2009.02.004.
doi:10.1007/978-3-540-30586-6\_88. [19] K. A. Rogova, K. Demuynck, D. V. Compernolle, Au-
[4] J. P. Kincaid, L. R. P. F. Jr., R. L. Rogers, B. S. Chissom, tomatic syllabification using segmental conditional
Derivation of new readability formulas (Automated random fields, in: Computational Linguistics in the
Readability Index, Fog Count and Flesch Reading Netherlands Journal, volume 3, 2013, pp. 34–48.
Ease formula) for Navy enlisted personnel, Re- [20] L. P. Dinu, V. Niculae, O. Sulea, Romanian syllab-
search Branch Report, Millington, TN: Chief of ication using machine learning, in: I. Habernal,
Naval Training, 1975. V. Matousek (Eds.), Text, Speech, and Dialogue -
[5] G. Marco, J. de la Rosa, J. Gonzalo, S. Ros, 16th International Conference, TSD 2013, Pilsen,
E. González-Blanco, Automated Metric Analysis of Czech Republic, September 1-5, 2013. Proceedings,
Spanish Poetry: Two Complementary Approaches, volume 8082 of Lecture Notes in Computer Science,
IEEE Access 9 (2021) 51734–51746. Springer, 2013, pp. 450–456.
[6] A. M. Ciobanu, L. P. Dinu, On the romanian [21] J. Krantz, M. W. Dulin, P. D. Palma, Language-
rhyme detection, in: Proceedings of COLING 2012: Agnostic Syllabification with Neural Sequence La-
Demonstration Papers, 2012, pp. 87–94. beling, 2019 18th IEEE International Conference
[7] L. Hjelmslev, The syllable as a structural unit, in: On Machine Learning And Applications (ICMLA)
the Proceedings of the 3rd International Congress (2019) 804–810.
of Phonetic Sciences (Ghent), 1938, volume 266, [22] V. N. Vitale, L. Schettino, F. Cutugno, On incre-
1938. menting interpretability of machine learning mod-
[8] F. De Saussure, Course in general linguistics, els from the foundations: A study on syllabic speech
Columbia University Press, 2011. units, in: F. Boschetti, G. E. Lebani, B. Magnini,
[9] D. Russo, The Notion of Syllable across History, N. Novielli (Eds.), Proceedings of the 9th Italian
Theories and Analysis, Cambridge Scholars Pub- Conference on Computational Linguistics, Venice,
lishing, 2016. Italy, November 30 - December 2, 2023, volume 3596
[10] M. Loporcaro, Syllable, segment and prosody, in: of CEUR Workshop Proceedings, CEUR-WS.org, 2023.
The Cambridge history of the Romance languages, URL: https://ceur-ws.org/Vol-3596/paper51.pdf.
2011, pp. 50–108. [23] O. Sulea, L. P. Dinu, B. Dumitru, Full inflec-
[11] W. Meyer-Lübke, Grammaire des langues romanes, tion learning using deep neural networks, in:
volume 4, H. Welter, 1906. A. F. Gelbukh (Ed.), Computational Linguistics
[12] M.-D. Glessgen, Linguistique romane: domaines and Intelligent Text Processing - 19th Interna-
et méthodes en linguistique française et romane, tional Conference, CICLing 2018, Hanoi, Vietnam,
Armand Colin, 2007. March 18-24, 2018, Revised Selected Papers, Part
[13] F. S. Miret, Fonética histórica, in: Manual de lingüís- I, volume 13396 of Lecture Notes in Computer Sci-
tica románica, Ariel España, 2007, pp. 227–250. ence, Springer, 2018, pp. 408–415. URL: https://doi.
[14] F. d’Ovidio, W. Meyer-Lübke, Grammatica storica org/10.1007/978-3-031-23793-5_33. doi:10.1007/
della lingua e dei dialetti italiani, volume 368, U. 978-3-031-23793-5\_33.
Hoepli, 1906. [24] M. Petrillo, F. Cutugno, A syllable segmentation al-
[15] G. Rohlfs, T. Franceschi, Grammatica storica della gorithm for english and italian., in: INTERSPEECH
lingua italiana e dei suoi dialetti: Morfologia, (No 2003, 2003, pp. 2913–2916.
Title) (1968). [25] Q. Dou, S. Bergsma, S. Jiampojamarn, G. Kondrak, A
[16] B. Bigi, C. Petrone, A generic tool for the automatic Ranking Approach to Stress Prediction for Letter-to-
syllabification of italian, A generic tool for the Phoneme Conversion, in: Proceedings of the Joint
automatic syllabification of Italian (2014) 73–77. Conference of the 47th Annual Meeting of the ACL
[17] C. R. Adsett, Y. Marchand, Are Rule-based Syl- and the 4th International Joint Conference on Nat-
labification Methods Adequate for Languages with ural Language Processing of the AFNLP: Volume 1 -
Low Syllabic Complexity? The Case of Italian, in: Volume 1, ACL ’09, Association for Computational
P. Wagner, J. Abresch, S. Breuer, W. Hess (Eds.), Linguistics, 2009, p. 118–126.
Sixth ISCA Workshop on Speech Synthesis, Bonn, [26] L. P. Dinu, An approach to syllables via
Germany, August 22-24, 2007, ISCA, 2007, pp. 58– some extensions of marcus contextual gram-
63. mars, Grammars 6 (2003) 1–12. URL: https://
[18] C. R. Adsett, Y. Marchand, V. Keselj, Syllabifi- doi.org/10.1023/A:1024089129146. doi:10.1023/A:
cation rules versus data-driven methods in a lan- 1024089129146.
guage with low syllabic complexity: The case of [27] L. P. Dinu, A. Dinu, On the data base of romanian
italian, Comput. Speech Lang. 23 (2009) 444– syllables and some of its quantitative and cryp-
tographic aspects, in: N. Calzolari, K. Choukri,
A. Gangemi, B. Maegaard, J. Mariani, J. Odijk,
D. Tapias (Eds.), Proceedings of the Fifth Interna-
tional Conference on Language Resources and Eval-
uation, LREC 2006, Genoa, Italy, May 22-28, 2006,
European Language Resources Association (ELRA),
2006, pp. 1795–1798.
[28] L. P. Dinu, A. Dinu, On the behavior of romanian
syllables related to minimum effort laws, in: Pro-
ceedings Workshop Multilingual Resources, Tech-
nologies and Evaluation for Central and Eastern
European Languages, co-located with RANLP 2009,
Borovets, Bulgaria 2006, 2009, pp. 9–13.
[29] S. Chebanow, On conformity of language structures
within the Indoeuropean family to poisson’s law,
Comptes rendus de l’Academie de science de l’URSS
55 (1947) 99–102.
[30] G. Altmann, Prolegomena to Menzerath’s Law,
Glottometrika 2 (1980) 1–10.
[31] A. M. Ciobanu, A. Dinu, L. P. Dinu, Predicting
romanian stress assignment, in: G. Bouma, Y. Par-
mentier (Eds.), Proceedings of the 14th Conference
of the European Chapter of the Association for
Computational Linguistics, EACL 2014, April 26-
30, 2014, Gothenburg, Sweden, The Association
for Computer Linguistics, 2014, pp. 64–68. URL:
https://doi.org/10.3115/v1/e14-4013. doi:10.3115/
V1/E14-4013.
[32] L. P. Dinu, A. M. Ciobanu, I. Chitoran, V. Nicu-
lae, Using a machine learning model to assess
the complexity of stress systems, in: N. Calzolari,
K. Choukri, T. Declerck, H. Loftsson, B. Maegaard,
J. Mariani, A. Moreno, J. Odijk, S. Piperidis (Eds.),
Proceedings of the Ninth International Conference
on Language Resources and Evaluation, LREC 2014,
Reykjavik, Iceland, May 26-31, 2014, European Lan-
guage Resources Association (ELRA), 2014, pp. 331–
336. URL: http://www.lrec-conf.org/proceedings/
lrec2014/summaries/1200.html.
[33] K. Cho, B. Van Merriënboer, C. Gulcehre, D. Bah-
danau, F. Bougares, H. Schwenk, Y. Bengio, Learn-
ing phrase representations using rnn encoder-
decoder for statistical machine translation, arXiv
preprint arXiv:1406.1078 (2014).
[34] D. P. Kingma, J. Ba, Adam: A method for stochas-
tic optimization, arXiv preprint arXiv:1412.6980
(2014).