Algorithm for Selecting Words to Compose Phonological Assessments

ao V

ıctor B. Marques

1 a

, Jo

ao Carlos D. Lima

1 b

, M

arcia Keske-Soares

2 c

, Cristiano C. Rocha

Fabr

ıcio Andr

e Rubin

and Raphael Vieira Miollo

Centro de Tecnologia, Universidade Federal de Santa Maria, Santa Maria, Brazil

Centro de Ci

encias da Sa

ude, Universidade Federal de Santa Maria, Santa Maria, Brazil

Xebia Data, Eindhoven, Netherlands

Petroleo Brasileiro S.A., Rio de Janeiro, Brazil

Keywords:

Speech Therapy, Phonological Processes, Graphs and Phonological Assessments.

Abstract:

The phonological assessment is one of the main resources that speech-language therapist has to identify phono-

logical disorders in children. For this, it is necessary to be composed of a words set that have a variety of

phonemes in different positions of the syllable and the word, in order to obtain a representative sample of

the phonological system. Thus, the present work aimed to analyze a set of 84 words from a phonological as-

sessment instrument, with the objective of identifying and removing words with over-represented phonemes.

Aiming to facilitate the phonological evaluation by making it more succinct with the reduction of the number

of words, the present work describes a judicious method organized in three steps, which was implemented in

Javascript and obtained a subset of 55 words, which have at least two occurrences of the same phonemes in

the proper positions in which they appeared in the initial set, representing a 35% reduction in the number of

words without losing quality.

1 INTRODUCTION

A thorough and comprehensive phonological assess-

ment is one of the main tools for the speech therapist

(Savoldi, 2012), as it helps in the identiﬁcation and

diagnosis of speech disorders. To compose a phono-

logical assessment tool, it is necessary to select words

that represent ﬁgures known to children, and that are

inserted in their basic vocabulary and social context

(Gomes et al., 2013).

In southern Brazil, the Child Phonological Assess-

ment (CPA) (Yavas et al., 2001) is one of the evalua-

tions used by speech therapists, and according to the

authors, 125 words were chosen, which represent the

vocabulary of children aged 3 with a balanced sam-

ple of the adult phonological system, and present, at

least, three possibilities of occurrence for each conso-

nant sound, in all possible syllable positions.

In this sense, in order to represent the adult phono-

logical system, it is important that the set of words is

https://orcid.org/0009-0007-3206-725X

https://orcid.org/0000-0001-9719-3205

https://orcid.org/0000-0002-5678-8429

comprehensive in relation to the variety of phonemes

and, also, that they are evaluated more than once

in different syllable and word positions (Pagliarin,

2009). Thus, it is also possible to evaluate and iden-

tify the phonemes present in the child’s phonetic in-

ventory, that is, those that he can reproduce sponta-

neously (Stoel-Gammon, 1985).

The selection of the best words to compose a

phonological assessment has been the subject of stud-

ies such as (Savoldi et al., 2013), where 116 words

were selected from an initial set of 722, after valida-

tion with experts in the ﬁeld. This number was re-

duced to 84 in the study of (Ceron et al., 2020), when

obtaining a balanced sample of words in an attempt to

avoid the over-representation of one phoneme or the

under-representation of another. However, it is ob-

served that many phonemes in this set still repeat in

the same position, with a frequency greater than the

minimum necessary for a quick and effective assess-

ment.

It was then sought to develop a method to eval-

uate a set of words through their phonetic transcrip-

tions, in order to determine the smallest sub-set that

contains a minimum number of occurrences for each

Marques, J., Lima, J., Keske-Soares, M., Rocha, C., Rubin, F. and Miollo, R.

Algorithm for Selecting Words to Compose Phonological Assessments.

DOI: 10.5220/0011990500003467

In Proceedings of the 25th International Conference on Enterprise Information Systems (ICEIS 2023) - Volume 1 , pages 80-88

ISBN: 978-989-758-648-4; ISSN: 2184-4992

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

phoneme in the different positions of the syllable and

the word, respecting the due restrictions that make a

reliable phonological assessment instrument.

The work is organized as follows. In the next sec-

tion, the basic concepts and related works that served

as the basis for this research will be presented. In

Section 3, the logic behind the choice of words that

make up a phonological assessment instrument will

be presented, and then the algorithm will be detailed.

In Section 4, the results obtained by this work will

be presented and discussed. Finally, in Section 5, the

conclusion is presented, and in the appendix all the

words of the initial set are found, and the crossed-out

words signal that these have been removed from the

ﬁnal set.

2 BACKGROUND AND RELATED

WORK

In this section, some theoretical concepts that underlie

the use of structures such as graphs in this study will

be addressed, as well as concepts from speech therapy

such as Phonological Processes (PPs) and how they

are connected to the present work.

2.1 Important Concepts

The reader will be brieﬂy introduced to some terms

from speech therapy, that are important for under-

standing the present work. In cases of phonologi-

cal disorders, only consonant phonemes are observed

(Shriberg et al., 1997). In addition, they can appear

in different positions in the syllable and word (begin-

ning, middle, and end), and the production of each

phoneme in a pronunciation must necessarily be in

one of these positions seen below.

• (OI) Onset Initial: beginning of syllable, word be-

ginning - ca.sa [house];

• (OM) Onset Medial: beginning of syllable, mid-

dle of the word - ca.va.lo [horse];

• (CM) Coda Medial: end of syllable, middle of the

word - ca.dar.c¸o [shoelace];

• (CF) Coda Final: end of syllable, end of the word

- a.mor [love];

• (OCI) Onset Complex Initial: beginning of sylla-

ble, beginning of word - Bra.sil [Brasil];

• (OCM) Onset Complex Medial: beginning of syl-

lable, middle of the word - bi.blio.te.ca [library].

2.2 Graphs

A graph G = (V, E) is a structure in which V is a

ﬁnite and non-empty set of n vertices, and a set E

of m edges, which are pairs of vertices of V. They

are classiﬁed according to the nature of the connec-

tion between their vertices, being able to be “undi-

rected” when their weights do not have direction, or

“directed”, as seen in Figure 1.

(a) Directed

e1:E1

e5:E5

e3:E3

e2:E2

e4:E4

(b) Double Directed

Figure 1: Types of Graphs.

In this work, a Double Directed Graph structure

was used, where each edge has, necessarily, two

weights: one for each direction. By analyzing the ex-

ample of Figure 2, we can notice that, given the vertex

p(OI) the edge that connects it to the vertex t(OM) has

different weights depending on the choice of the cen-

tral vertex, highlighting the importance of centrality

in graphs to determine the inﬂuence of one vertex on

another.

d(OCM)

p(OI)

ɾ(OCM)

s(CM)

ɾ(CM)

t(OM)

ɾ(OM)

s(OM)

2:4

2:13

4:10

4:6

6:10

4:15

4:3

13:4

4:6

Figure 2: Partial mapping of the entry data.

The weights of each vertex are determined ac-

cording to the number of words in the set in which

a phoneme can occur in a certain position. It is also

worth of notice that words which have more than one

phoneme inﬂuence the weight of more than one ver-

tex. Therefore, when removing or adding words to

a vertex, the weights of its edges will be updated as

well.

Algorithm for Selecting Words to Compose Phonological Assessments

2.3 Graph Centrality

Introduced by (Bavelas, 1948) while studying the

communication of individuals and inﬂuence in small

social groups, the concept of centrality in graphs is as-

sociated with the degree of importance and inﬂuence

that each vertex has on another in the graph, and what

bottlenecks may exist in their connections. (Freeman,

1978) also works with the same concepts of central-

ity in social networks, investigating the quantitative

measures capable of deﬁning the importance of each

vertex.

In this context, this work seeks to identify the im-

portance that each phoneme has in the set of analyzed

words, considering possible phonetic transcriptions

where a phoneme can appear in more than one po-

sition of the syllable and the word. In this way, it is

possible to identify the most inﬂuential phonemes in

the set, that is, those that are over-represented, and

make them less important by removing some of their

words, making the set more balanced.

Each vertex has a list of words in which a

phoneme occurred in a certain position. With that, it is

possible to identify the importance that each word has

in the vertices, in order to list which could or could not

be removed from the graph without the same ending

with under-represented phonemes.

2.4 Phonological Processes

In the context of speech therapy, phonological pro-

cesses have a great inﬂuence on a child’s language

acquisition process. It is expected that during this

stage, she applies several phonological processes,

such as replacing one phoneme with another or omit-

ting them. Such substitutions and omissions are con-

sidered in speech therapy as Phonological Processes

(PP), and some examples are presented in Table 1.

However, if a PP persists for a long time, it can

become a phonological disorder and remain in the

child’s speech, accompanying her in school during

her literacy process, bringing harm to her social life

(Goulart and Chiari, 2014). Therefore, some works

are dedicated to identifying possible phonological

disorders through voice recognition in phonological

evaluations (Franciscatto et al., 2019), so that the di-

agnosis and treatment is given early.

In the work of (Franciscatto et al., 2019), Ma-

chine Learning (ML) techniques were used to classify

the pronunciations of 84 words as correct or incor-

rect and to recognize phonological processes through

them. In (Franciscatto. et al., 2018), a case-based

method, commonly used in the health ﬁeld and in ML

techniques (Tavana et al., 2022), is developed and is

able to have good learning while allowing new cases

to be stored in a database without complications (Hu-

sain and Pheng, 2010). The method works as an ex-

tra validation layer after the pronunciation classiﬁca-

tion by ML, registering new cases and validating them

with an expert.

However, despite studies using Artiﬁcial Intelli-

gence (AI) techniques (Iliya and Neri, 2016) and ML

(Franciscatto et al., 2019) in speech therapy, they are

only used to identify phonological processes and clas-

sify (correct/incorrect) the pronunciation of spoken

words in phonological evaluation. But, before that,

a set of words must be chosen by a specialist to be

pronounced by the children, and such a set must an-

alyze all the phonemes of the language in different

positions of the word. Thus, the choice of words in

the set must follow speciﬁc criteria, addressed in Sec-

tion 3.1, because in a phonological evaluation, for ex-

ample, all phonemes must be analyzed at least twice

(Stoel-Gammon, 1985). So, to deﬁne the smallest

subset of words that meets the same criteria as the ini-

tial set, it is a matter of quantifying the “least effort”,

discussed in Section 3, rather than learning from er-

rors and successes.

3 DEVELOPMENT

In this section, the operation of the algorithm will be

presented and detailed, from the basic input structures

to the ﬁnal result.

The set of 84 words from (Ceron et al., 2020) was

used as the database. Additionally, all the words in

the set should be phonetically decomposed in order to

detail in which positions the phonemes that compose

them appear. For this, the JSON structure, developed

in the work of (Marques, 2022), was used, which is

synthesized in Figure 3.

Figure 3: Example of the word “cavalo [horse]” [horse] de-

composed in consonant phonemes.

The main idea of the algorithm is to avoid as much

as possible that the same phoneme is over-represented

in the set of words, being analyzed in the same posi-

tion with a frequency equal to or greater than the min-

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

Table 1: Phonological processes in Portuguese acquisition. (Yavas et al., 2001).

Phonological Process (PP) Deﬁnition Example

Cluster reduction Reduction of a consonant within the same syllable. bruxa (witch) [br

usa]→[b

usa]

Final fricative deletion Deletion of phoneme /s/ in both syllable and word ﬁnal positions l

apis (pencil) [l

apis]→[

api]

Liquid substitution Substitution of one liquid for another. zero (zero) [zeru]→[selu]

Plosivation Substitution of a fricative consonant for a plosive vaca (cow) [vaka]→[baka]

Intervocalic liquid nasalization Substitution of a liquid by a nasal in intervocalic position carro (car) [k

aχ u]→[k

amu]

imum necessary, avoiding the under-representation as

well.

In order for a phonological assessment to be reli-

able and the obtained set can be used, it is necessary

to follow some rules of speech therapy, which will be

discussed in Section 3.1.

3.1 Phonological Assessment Rules

In order for an assessment to be reliable and compre-

hensive, phonemes need to be analyzed in different

positions of the syllable and word with a deﬁned min-

imum frequency, in order to contain a balanced sam-

ple of the adult phonological system (Savoldi, 2012).

Also, to determine the presence or absence of the

sound in the phonetic inventory, a minimum of two

occurrences of the segment can be considered, re-

gardless of the position in the word (Stoel-Gammon,

1985). Studies such as (Yavas et al., 2001) consider,

for phonological assessment, a minimum of three pos-

sibilities of occurrence for each consonant phoneme

in all possible syllable positions. In this work, a min-

imum of two occurrences of the same phoneme eval-

uated in a certain position was considered, according

with (Stoel-Gammon, 1985). Aiming at the ﬂexibil-

ity of the algorithm for its applicability in different

scenarios from this research, the minimum number of

occurrences is one of the inputs provided by the user.

It should be noted that not all phonemes appear

in all allowed positions, due to a limitation of the

language itself (Savoldi, 2012). In the input data, it

was observed that the phoneme *n always appears in

word productions in the diminutive form in OM and

therefore should not have an impact on the proposed

method, as it would not occur in words in their natural

form.

In short, the method introduced by this work must

follow the criteria:

C.1 The phoneme /η/ evaluated in OM is disregarded

— words in diminutives;

C.2 Each phoneme must continue to occur at least

2 times in the same position after deleting any

word.

Next, in Section 3.2 the ﬁrst step of the algorithm

will be detailed.

3.2 Phoneme Mapping and Graph

As a ﬁrst step, the words that have a certain phoneme

in a certain position are counted on a map. The pair

“phoneme (position)” will be the key of each entry in

the map, which will point to a list of words in which

the phoneme appears in the right position. As a result,

a graph structure is obtained, shown in Figure 4.

p(OI)

passarinho [bird]

pastel [pastry]

pedra [rock]

porta [door]

s(OM)

passarinho [bird]

bolsa [handbag]

(+4)

ɾ(OM)

passarinho [bird]

fogo [fire]

(+8)

t(OM)

porta [door]

pastel [pastry]

(+13)

d(OCM)

pedra [rock]

vidro [glass]

ɾ(OCM)

chifre [horn]

pedra [rock]

(+11)

ɾ(CM)

garfo [fork]

jornal [newspaper]

porta [door]

s(CM)

plástico [plastic]

pastel [pastry]

(+4)

Figure 4: Sample of the graph generated after phoneme

mapping.

The algorithm must receive a set of words bro-

ken down into their consonant phonemes, as shown

in Figure 3. In this study, the abstract set from the

work of (Marques, 2022) was used as input, which

decomposed the 84 words proposed by (Ceron et al.,

2020) into consonant phonemes with the validation of

experts in the ﬁeld.

From the analysis of Figure 4, we observe that

some phonemes (/t/) occur in the same position in a

greater number of words (15) than the minimum nec-

essary established by criterion C.1. Therefore, since

the phoneme repeats a lot in the set, it is thought

that some words can be excluded from the evaluation,

since they would be generating a super-representation

of the phoneme.

Algorithm for Selecting Words to Compose Phonological Assessments

But which ones could be excluded? It is noted

that the phoneme /s/ occurs in Onset Medial (OM)

in 6 words. According to the minimum value of oc-

currences criterion, 4 of these words could be disre-

garded, but it is not that simple. Before removing a

word from the evaluation, it is necessary to remember

that it is composed of other phonemes that are evalu-

ated in different positions. And, if the word that we

are willing to exclude is one of the few that evaluates

some other phoneme in a certain position? All these

issues are considered by the algorithm proposed by

this work.

3.3 Word Analysis

As a second step, through the graph shown in Fig-

ure 4, we will analyze each node and the words con-

tained in it. As we visit each node, we can remove

words, as long as the graph remains valid according

to the criteria established in Section 3.1.

Given a node of the graph, we need to determine

if the phoneme is over-represented in a certain posi-

tion or not. As an example, we will work with the

node highlighted in Figure 4, where the phoneme /p/

appears in OI in 4 different words. Thus, it can be

said that this pair is over-represented, since it occurs

in more words than the minimum necessary.

Now, we need to check if it is possible to remove

any of the words, without compromising the ﬁnal set

with under-represented phonemes. We do this by an-

alyzing each word of the visited node separately, as

shown in Figure 5.

p(OI)

passarinho [bird]

pastel [pastry]

pedra [rock]

porta [door]

s(OM)

ɾ(OM)

t(OM)

d(OCM)

ɾ(OCM)

ɾ(CM)

s(CM)

Weight of the Edges

of Each Word

s(OM) 6

ɾ(OM) 10

s(CM) 6

t(OM) 15

d(OCM) 2

ɾ(OCM) 13

ɾ(CM) 3

t(OM) 15

Importance

Ordering the Words

according to their Weights

2 pedra [rock]

3 porta [door]

6 passarinho [bird]

6 pastel [pastry]

10 passarinho [bird]

13 pedra [rock]

15 pastel [pastry]

15 porta [door]

passarinho

[bird]

pastel

[pastry]

pedra

[rock]

porta

[door]

Figure 5: Analysis and ordering of each word of the visited

node.

The calculation of the weights of each word is

done through the counting of the weight of each node

in the remaining graph in which it appears. In Fig-

ure 4, the word “passarinho [bird]” appears in the

nodes “s(OM)” and “l(OM)”, which have weights

of 6 and 10, respectively. With the ordering of the

weights, we can evaluate the importance of each word

in that node, and it would be enough to keep the

ﬁrst two and exclude the others, while still keeping

the graph valid. However, when removing words,

it is important to be careful not to exclude any that

have a weight lower than the established minimum,

as we will impact other nodes, and these may become

under-represented. Figure 6 shows the steps followed

so far, with the removal of words and subsequent up-

dating of the graph, ﬁnishing the ﬁrst cycle of the al-

gorithm.

To better visualize the impact that the ﬁrst cycle of

exclusion had on the graph, we can check in Figure 7

the updated weight of each node after the removal of

words.

However, the algorithm does not have a rule for

choosing the next node to be visited, being chosen

by order of insertion into the graph’s data structure.

In each visited node, words can be removed and the

graph must be updated, making the algorithm possi-

bly reproduce different results depending on the or-

dering of the graph.

4 RESULTS AND VALIDATION

All the stages of the algorithm described in Section 3

were implemented in the Javascript language. As in-

put for the method, a set of correct phonetic transcrip-

tions in JSON of the 84 words proposed by (Ceron

et al., 2020) was used. After all the data in the set

were analyzed by the algorithm, a reduction of 29

words was obtained, resulting in a subset of 55 words,

which satisﬁes the criteria established in Section 3.1.

However, to validate how the reduction of words

proposed by this work would behave in a real sce-

nario, data from 1611 phonological evaluations ap-

plied to 1357 children from April/2013 to Jan-

uary/2017 were analyzed. Some evaluations were in-

complete in the database used, so that some words

were not spoken in certain evaluations. For this rea-

son, it was decided to consider only evaluations with

42 or more words spoken by the children, which rep-

resent half of the initial set of words, resulting ﬁnally

in 1587 evaluations as a base for validation.

The PCC-R index (Percent of Consonants

Correct-Revised), developed by (Shriberg et al.,

1997), was used as a metric to validate the new

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

Choose Node

not Visited

Weights of the

Words in Node

Ordering the Words

by their Weights

Reorganize the Node

according with weights

Remove most

repeated words

p(OI)

passarinho [bird]

pastel [pastry]

pedra [rock]

porta [door]

Weight of the Edges of

Each Word

s(OM) 6

ɾ(OM) 10

s(CM) 6

t(OM) 15

d(OCM) 2

ɾ(OCM) 13

ɾ(CM) 3

t(OM) 15

Ordering the Words

according to their Weights

2 pedra [rock]

3 porta [door]

6 passarinho [bird]

6 pastel [pastry]

10 passarinho [bird]

13 pedra [rock]

15 pastel [pastry]

15 porta [door]

p(OI)

pedra [rock]

porta [door]

passarinho [bird]

pastel [pastry]

Update graph

End

Have all nodes been visited?

Yes

Set the current

node as Visited

Figure 6: Flowchart of the analysis of each phoneme represented in the graph.

p(OI)

pedra [rock]

porta [door]

s(OM)

travesseiro [pillow]

bolsa [handbag]

(+3)

ɾ(OM)

travesseiro [pillow]

fogo [fire]

(+7)

t(OM)

porta [door]

batom [lipstick]

(+12)

d(OCM)

pedra [rock]

vidro [glass]

ɾ(OCM)

chifre [horn]

pedra [rock]

(+11)

ɾ(CM)

garfo [fork]

jornal [newspaper]

porta [door]

s(CM)

plástico [plastic]

floresta [forest]

(+3)

Figure 7: Graph sample after removing two words from the

visited node.

dataset. This index is also used as a basis for recom-

mending therapeutic activities in the system proposed

by (Franciscatto et al., 2021). The PCC-R value is

calculated using Equation 1, based on the number of

correct consonants (CC) produced by the child and

the total expected productions (TEP). This allows us

to determine the severity of the phonological disor-

der, as shown in Table 2. Additionally, it is directly

related to the presence of phonological processes in

the child’s speech, as found in the study by (Ghisleni

et al., 2010).

PCC-R =

T EP

× 100 (1)

Table 2: Indication of speech disorder according with PCC-

R value (Shriberg et al., 1997).

PCC-R Value Indication of Disorder

Less than 50% High

Between 51% e 65% Moderate-High

Between 66% e 85% Low-Moderate

Greater than 85% Low

Thus, the PCC-R of each assessment was calcu-

lated by ﬁrst considering all 84 words of the initial

set. Then, another PCC-R was calculated in the same

way, but this time considering only the 55 words of

the new proposed set. The goal was to determine if

the degree of disorder associated with the ﬁrst PCC-R

remained in the second case. The result of this calcu-

lation can be seen in the graph of Figure 8.

Figure 8: Comparative of PCC-R Classiﬁcations.

Algorithm for Selecting Words to Compose Phonological Assessments

However, in order to identify if there were changes

in the disorder classiﬁcation of an assessment when

only the 55 words of the proposed subset were con-

sidered, the two PCC-R results in each assessment

were compared, and each change was calculated. Af-

ter that, we arrived at the graph shown in Figure 9,

where it is possible to identify that 96% of the as-

sessments had no change in the Degree of Disorder

associated with the PCC-R when the new subset was

considered.

96,0%

Figure 9: Changes in Indication of Speech Disorder when

only the 55 selected words were considered.

Nevertheless, 64 evaluations, which represent 4%

of the evaluations analysed, had changes in the clas-

siﬁcation of the Degree of Disorder. Such changes

are mapped in the graph of Figure 10, in which it is

possible to identify that the changes are mainly con-

centrated in lower degrees.

Figure 10: Changes in Classiﬁcation of 64 assessments.

5 CONCLUSION

This work aimed to perform a computational analy-

sis of the 84 words proposed by the work of (Ceron

et al., 2020) to compose a phonological assessment

tool. The analysis focused on identifying words that

generate a super-representation of their phonemes in

the set and to perform a critical removal of them. To

do this, it was necessary to establish rules for the val-

idation of removals, in order to not generate under-

represented phonemes in the set.

The method was described in stages in Section 3,

and the algorithm was implemented in Javascript and

validated with the use of real data in Section 4. The al-

gorithm was efﬁcient in reducing the number of words

used by speech therapists in phonological evaluations

of children, also representing a reduction in the spent

time applying the evaluations. In addition, the reduc-

tion of the words did not bring negative impacts to the

quality of the evaluation, since the algorithm was im-

plemented following pre-established criteria in Sec-

tion 3.1. The validation was done using the PCC-R

value of (Shriberg et al., 1997), which indicates the

degree of phonological disorder based on the number

of correct consonants produced in the evaluation.

The set proposed by (Ceron et al., 2020) was used

as input for the algorithm, and at the end of the anal-

ysis of the 84 initial words, it was identiﬁed that 55

of them would already be sufﬁcient to analyze the

same phonemes in the same positions that the ini-

tial set evaluated, with at least 2 occurrences for each

phoneme in each position, representing a reduction of

35% in the set size. Therefore, this work proposes

a method that can be followed manually or imple-

mented in any programming language to analyze the

words that compose a phonological assessment tool,

with the aim of making it more succinct without los-

ing quality.

REFERENCES

Bavelas, A. (1948). A mathematical model of group struc-

tures. Human Organization, 7:16–30.

Ceron, M. I., Gubiani, M. B., Oliveira, C. R. d., and Keske-

Soares, M. (2020). Phonological assessment instru-

ment (infono): A pilot study. CoDAS, 32(4).

Franciscatto, M. H., Del Fabro, M. D., Damasceno Lima,

J. C., Trois, C., Moro, A., Maran, V., and Keske-

Soares, M. (2021). Towards a speech therapy support

system based on phonological processes early detec-

tion. Computer Speech & Language, 65:101130.

Franciscatto, M. H., Lima, J. a. C. D., Trois, C., Maran,

V., Soares, M. K., and Rocha, C. C. d. (2019). Ap-

plying situation-awareness for recommending phono-

logical processes in the children’s speech. SAC ’19,

page 739–746, New York, NY, USA. Association for

Computing Machinery.

Franciscatto., M. H., Lima., J. C. D., Moro., A., Maran.,

V., Augustin., I., Keske Soares., M., and Cortez da

Rocha., C. (2018). A case-based system architecture

based on situation-awareness for speech therapy. In

Proceedings of the 20th International Conference on

Enterprise Information Systems - Volume 1: ICEIS,,

pages 461–468. INSTICC, SciTePress.

Freeman, L. C. (1978). Centrality in social networks con-

ceptual clariﬁcation. Social Networks, 1(3):215–239.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems

Ghisleni, M. R. L., Keske-Soares, M., and Mezzomo, C. L.

(2010). The use of repair strategies considering the

severity of the evolutional phonological disorder. Re-

vista CEFAC: Atualizac¸

ao Cient

ıﬁca em Fonoaudiolo-

gia e Educac¸

ao, 12(5):766–771.

Gomes, A. M., Feltin, T. D., Savoldi, A., and Keske-Soares,

M. (2013). Phonological assessments: analysis of the

most frequent words considering the vocabulary and

sound class. Dist

urbios da Comunicac¸

ao, 25(1).

Goulart, B. N. G. d. and Chiari, B. M. (2014). Speech

disorders and grade retention in elementary. Revista

CEFAC: Atualizac¸

ao Cient

ıﬁca em Fonoaudiologia e

Educac¸

ao, 16(3).

Husain, W. and Pheng, L. T. (2010). The development of

personalized wellness therapy recommender system

using hybrid case-based reasoning. In 2010 2nd In-

ternational Conference on Computer Technology and

Development, pages 85–89.

Iliya, S. and Neri, F. (2016). Towards artiﬁcial speech

therapy: A neural system for impaired speech seg-

mentation. International Journal of Neural Systems,

26(06):1650023. PMID: 27354188.

Marques, J. V. B. (2022). M

odulo de an

alise contrastiva

para avaliac¸

oes fonol

ogicas de e-fono.

Pagliarin, K. C. (2009). A abordagem contrastiva na ter-

apia fonol

ogica em diferentes gravidades do desvio

fonol

ogico. Mestrado em dist

urbios da comunicac¸

humana, Universidade Federal de Santa Maria, Santa

Maria.

Savoldi, A. (2012). Instrumento de avaliac¸

ao fonol

ogica:

validac¸

ao de conte

udo. Mestrado em dist

urbios da

comunicac¸

ao humana, Universidade Federal de Santa

Maria, Santa Maria.

Savoldi, A., Ceron, M. I., and Keske-Soares, M. (2013).

What are the best words to compose an evaluation

phonological instrument? Audiology-Communication

Research, 18(3):194–202.

Shriberg, L. D. et al. (1997). The speech disorders classi-

ﬁcation system (sdcs). Journal of Speech, Language

and Hearing Research, 40(4):723–740.

Stoel-Gammon, C. (1985). Phonetic inventories, 15-24

months: a longitudinal study. Journal of Speech and

Hearing Research, 28:505–512.

Tavana, M., Nazari-Shirkouhi, S., Mashayekhi, A., and

Mousakhani, S. (2022). An integrated data mining

framework for organizational resilience assessment

and quality management optimization in trauma cen-

ters. SN Operations Research Forum, 3:17.

Yavas, M., Hernandorena, C., and Lamprecht, R. (2001).

Avaliac¸

ao fonol

ogica da crianc¸a: reeducac¸

ao e ter-

apia. Biblioteca Artmed. Fonoaudiologia. Artmed Ed-

itora.

Algorithm for Selecting Words to Compose Phonological Assessments

APPENDIX

Table 3: Phonemes and words in which it occurs in a certain position in the initial set. Scratched words are not in the new set.

ICEIS 2023 - 25th International Conference on Enterprise Information Systems