Lex-Libras: Morphosyntactic Model of the Brazilian Sign Language to
Support a Context-based Machine Translation Process
onio M. Silva
1 a
, Tanya A. Felipe
2 b
, Laura S. Garc
3 c
, Diego R. Antunes
4 d
and Andr
e P. Guedes
3 e
Engineering Department, State University of Par
a, Bel
em, PA, Brazil
National Institute for Deaf Education, Rua das Laranjeiras, 232, Rio de Janeiro, RJ, Brazil
Department of Informatics, Federal University of Paran
a, Curitiba, PR, Brazil
Informatics Department, Federal University of Technology, Ponta Grossa, PR, Brazil
Lex-Libras, Morphosyntactic Rules, Translation, Brazilian Sign-language.
Brazilian Sign Language (BSL–Libras) is the preferential language of Deaf communities in Brazil. The
Human-Computer Interaction Architecture in Sign Language (HCI-SL) was proposed, which will offer
user-system interaction in BSL. In 2015 this architecture had its formal model developed, together with
the phonological sign decomposition. The work described here advanced proposing morphological rules.
Differently from American Sign Language, Libras has a group of verbs that inflect, so translators need to
represent this phenomena in the automatic generation process and reflect it in its output, the 3D Avatar.
Related Portuguese-BSL translators available have not yet been able to generate correct BSL sentences with
regard to these verbs. This paper presents the Lex-Libras modeling process, a set of rules in the form of
a Context-Free Grammar capable of describing the morphosyntactic level of BSL. These rules compose the
morphosyntactic model of architecture in the Brazilian Portuguese to Brazilian Sign Language semiautomatic
translation process through an avatar.
Deaf people in many countries have their own
Sign Language (SL) as part of their cultures.
The SL are the natural and preferred languages
of deaf communities in several countries (Felipe,
2006), and many of them are also used by
Indigenous communities in which there are deaf
Indians. However, although the SL are also inserted
in other cultures predominantly of listeners who
use auditory-oral languages, the Deafs still face
difficulties that often relate to the processes of
teaching and learning and with computational tools
that could be an alternative to true school inclusion
and citizenship.
For (Ferreira and Garc
ıa, 2018), an accessible
environment to deaf students should offer SL
interpreted in similar ways to those expressed in
the real context. There are some related translation
services from Brazilian Portuguese (PT-BR) to BSL
available in (Ara
ujo, 2012), (Lima, 2015), (Felipe,
2013) and (Felipe, 2014) - TLibras whose objective
was building and intelligent Libras avatar
, and (De
Martino et al., 2017), but they still do not present an
adequate treatment for certain grammatical aspects of
BSL what leads to the generation of ungrammatical
Built in order to represent Sign Language in real
context, the HCI-SL (Garcia et al., 2013), seeks
to offer tools that enable the interaction of deaf
people with the technology in SL. In addition, the
formal computational model for the representation of
sign languages (CORE-SL) (Antunes, 2015) stated a
set of linguistic-computational requirements for the
proper functioning of the architecture. Continuing
This adjective is being used to differentiate avatars as
computational animation, from the avatar that will generate
signals from a linguistic knowledge of two languages
with equivalence rules from a semi-automatic translation
Silva, A., Felipe, T., García, L., Antunes, D. and Guedes, A.
Lex-Libras: Morphosyntactic Model of the Brazilian Sign Language to Support a Context-based Machine Translation Process.
DOI: 10.5220/0010442402670274
In Proceedings of the 23rd International Conference on Enterprise Information Systems (ICEIS 2021) - Volume 2, pages 267-274
ISBN: 978-989-758-509-8
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
this research path, the authors of the present paper
built the morphosyntactic rules that will solve the
grammatical problems found in related literature.
The main problem this research aimed to solve was
the context-dependency of the Sign Language. An
example of this dependency is the sign for the verb
”fall”, which varies according to the sentence subject.
The aim of this article is therefore to present
the Lex-Libras, a rule-based transcription model
that establishes the link between phonological and
morphosyntactic levels. This model can provide
context-based Sign Language generation and, in so,
more quality and assertiveness to automatic and
semi-automatic translation processes. As a result,
in addition to contributing to the translation process
that considers the context, Lex-Libras proved the
hypothesis of (Antunes, 2015) that it is possible
to use the phonological level as an input to the
morphological level through a bottom-up approach
The aim of the TLibras Project (Felipe, 2003) was
to develop a translator from PT-BR to BSL. It
also aimed at the creation of phonological rules
for the generation of signals. However, the team
responsible for creating the avatar and programming
the phonological level for the inclusion of signals
could not finish this step. Therefore, the team that
worked with the morphosyntactic level, created a
translation process that was named PUL0 - (Oralizer
UNL-LIST from Portuguese) (Felipe, 2003) a system
of automated and unidirectional translation from
an oral-auditive language, PT-BR, into the linear
representation (Libras Script for Translation – LIST)
of a gestural-visual language. Yet, the team
could not test this translation system. Its goal
was to convert a sentence originally produced in
PT-BR to a specialized transcription of BSL. That
would be signaled by through an intelligent avatar.
Such avatar would signal from the phonological
configurations, that is, from the five parameters of
contrast: hand configuration (HC), location (LOC),
hand arrangement (HA), directionalityy-contact (C),
and non-manual expressions (NME).
Another initiative was Prodeaf
, a text and
voice translation proposal from PT-BR to BSL.
The translation is made by two tools: a mobile
application and a plugin for websites. Currently,
In this approach, we start modeling at the lowest level
- the phonological level, to, then, move up to the highest
levels of the language.
HandTalk team
, which was purchased the Prodeaf
project. HandTalk was derived from a project
called Falibras, a computer system that converts texts
and audios into BSL. HandTalk performs machine
and semi-automated translation through a website
translator and an app.
Another translation proposal was the VLibras
ujo, 2012), which consists of a set of tools
aiming to translate digital content (text, audio, and
video). According to the research group working with
this application, the system presented syntactic and
semantic deficiencies, and, therefore, an automatic
translation component was developed, which brings
a formal language of syntactic-semantic translation
rules and also a set of grammar rules (Lima, 2015).
The research related to the morphosyntactic and
discursive issues of this avatar-translator is still
The research of (De Martino et al., 2017)
and (Paiva, 2019) presented a rule-based machine
translation system, which has a translator module that
analyzes the inputs and then converts them into an
intermediate representation, called an “intermediate
language” which serves as the input for the animation
A morphosyntactic evaluation of these translators
was presented by (Silva, 2020). The author, together
with a Brazilian Sign Language expert, considered an
PT-BR input and observed if the output was generated
according to the grammar of BSL. From the analysis
of these translators, it is possible to see that, although
they have achieved their objectives, they still generate
signs or phrases that are considered ungrammatical by
the deaf community. We hypothesize, therefore, that
such incongruity occurs due to the lack of inclusion
of phono-morphosyntactic-semantic-discursive
equivalence rules for the units in both languages.
The HCI-SL Architecture (Garcia et al., 2013) aims to
provide the development of an integrated environment
for the Deafs. Such an environment has been built
through methodological strategies and services that
can correctly solve, both from the linguistic and
computational points of view, issues related to the
computational treatment of SL and, consequently,
assist in the elimination of the social barrier of access
to information and knowledge suffered by the Deafs.
The rules presented in Lex-Libras is a new component
of the HCI-SL and will integrate its CORE-SL formal
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
model(Antunes, 2015), (Antunes et al., 2015).
For the CORE-SL project (Antunes, 2015), the
author stated a formal structure for computational
representation of the signals. This representation
considered all layers of the HCI-SL (Garcia et al.,
2013). In addition to the computational approach,
linguistic-computational aspects and requirements
were designed to provide adequate support to the
architecture. Through CORE-SL, it was possible to
create a set of sample signals which was more suitable
for training and evaluating the processes involved in
the translation. That is, the database can be built from
the parameters which were specified in the model,
enabling a small set of signals that can exemplify
several cases of SL through phonemes (Antunes et al.,
2015). Figure 1 shows the linguistic model, which
was extended with the Lex-Libras component through
the research presented in the next sections.
Figure 1: CORE-SL layered model.
4.1 Methodological Steps
For the development of the present research, we
followed these steps: 1. Study of the Morphology
of BSL, especially regarding the verbs and their
inflection systems. This research aimed at creating
a conceptual basis for the computational model. 2.
Review of Related Works to understand the context
in which the research is carried out. 3.Meetings
with a computational linguist specialized in BSL
for the Morphosynthatic Evaluation of the PT-BR
- BSL translators (Silva, 2020), to verify how the
programs generate lexical items from PT-BR input; 4-
Definition of the Computational Model in tree format;
5- Meetings with a computational linguist specialized
in BSL, for the evaluation of the proposed model; 6-
Adjustments in the model and definition in the format
of a Context-Free Grammar (CFG).
4.2 The Morphosyntactic Level
In BSL, as in any other natural language, the formal
aspects (phonological, morphological and syntactic
levels) that relate to the aspects of meaning (semantic
and pragmatic) are related or projected. In the SL, the
sign corresponds to the lexical item in oral-auditory
languages. Within its rules for the creation of new
signs, there is a combination of minimal units with
meaning (morphemes) with each other, a process that
gives rise to new signs.
According to (Felipe, 2006) the five phonological
parameters of BSL, hand configuration (HC),
location (LOC), hand arrangement (HA),
directionality-contact (C), and non-manual
features (NME) can be morphemes at the
morphosyntactic-discursive level. They can also
be classified into two types, as in oral-auditory
Lexeme: they are the units of lexical meaning,
the “roots”. In Brazilian Portuguese verbs: achar,
andar, amar (to buy, to walk, to love) - the lexemes
are {compr-} {and-} {am-};
: are the designations, prefixes,
suffixes, infixes and gender desinence: {-o}, {-a};
verbal endings: thematic vowels: {-a}, {-e}, {-i};
personal-number ending: {-o},{-s}, { }, {-mos},
{-is}, {-m};
Identifying these minimal components is useful for
computational processing, both in oral-auditory and
gestural-visual languages.
4.3 Verbal Inflection in BSL
Verbal inflection can occur through five processes:
inflection for the person of speech, inflection for the
verbal aspect (Finau, 2004), inflection for gender,
inflection for the locative, and verbs with frequency
inflection or modal cases (Felipe, 1998a), (Felipe,
1998b), (Felipe, 2002), (Felipe, 2006), (Felipe, 2007).
Inflection for the person of speech, inflection for
gender and inflection for the locative were selected
for this research. Felipe is the Brazilian Sign
Language researcher that adapted the American Sign
Language (Stokoe, 1960) system to the Brazilian Sign
Language, a language which, differently from the
American one, has verbal inflexion.
4.3.1 Person of Speech Inflection
Through displacement, the direction of the movement
parameter can make rectilinear or semi-circular
These examples of grammemes, refer to the Portuguese
Lex-Libras: Morphosyntactic Model of the Brazilian Sign Language to Support a Context-based Machine Translation Process
trajectories, thus inflecting the person of speech. The
beginning and the ending of the movement mark
the subject and object that agree with the verb.
The signaling of this verb is carried out through
the movement of hand displacement concerning the
person of speech.
The person will be signaled by the active hand in
direction to the referred person of speech, being the
speaker the first person. In the photo
, the speaker,
looking at the interlocutor, performs a movement
directed at this interlocutor, which will be transcribed
as ”1sPERGUNTAR2s”, translated: “I ask you”.
4.3.2 Gender Inflection
Gender inflection does not refer to male and female
aspects in BSL, but, as in some oral-auditory
languages, it is a classification for people, animals,
objects, plants, or vehicles (Felipe, 2002). In this
case, there is the animacy marking: animate (person
or animal) or inanimate (objects, plants or vehicle). In
classifier verbs, the HC parameter will be performed
as a grammeme. That is, as a mark of a verbal
agreement with the subject or object of the sentence,
depending on the verb
. In the example, the type of
object defines the form of the verb ”CAIR” (”fall”).
4.3.3 Locative Inflection
The LOC parameter can indicate a type of inflection.
the verb ”COLOCAR” (to put)
, which in addition to
being a classifier verb, also agrees for the locative.
The locative agreement mark indicates that the direct
object of the sentence will have its final point of
displacement in the locative presented prior to the
direct object. That is why the order is so relevant for
this type of verb. Thus, first is the locative sign, then
the object sign, and, finally, the verb sign that will end
its articulation at the location.
4.4 Conceptual Model
The unique CORE-SL architecture(Antunes et al.,
2015) was the insertion of the signal from the
phonological level. Though considering the
morphological level, CORE-SL did not include a
proposal for the connection with the morphological
and syntactic - discursive levels; therefore, the
modeling of the Lex-Libras complements this model.
Thus, the morphological model receives as input the
phonological model, defining a morpheme. Such
Figure is available in http://bit.ly/3pCkNto
Figure is available in https://bit.ly/3dqFXbr
Figure is available in http://bit.ly/3dvQW3w
morpheme will provide input to the syntactic model,
and, finally, the syntactic will provide input to the
phono-morphosyntactic-discursive model. Figure 2
presents the structure.
Figure 2: Language model for CORE-SL adapted from the
addition of the morphosyntactical rules.
4.5 Formal Model
The conceptual vision of Lex-Libras was modeled in
a hierarchical tree format to assist interdisciplinary
research (education, linguistics, computing, among
others). However, according to (Antunes, 2015), such
a view can lead to ambiguous descriptions, as they
do not present the rules or methods of how these
rules should be created. Thus, after the previous
definitions, the rules were formalized through a
CFG (Context-Free Grammar), a formalism widely
accepted and used for works of this nature (Amaral,
2012), (Antunes, 2015), (De Martino et al., 2017).
β α, where:
α consists of an arbitrary sequence of terminal or
non-terminal symbols;
β consists of a singular non-terminal symbol;
Thus, any occurrence of β during the parsing phase,
this non-terminal symbol can be replaced by a α
regardless of context. In the context of Computing,
the definition of a formal grammar is adopted through
an EBNF (Extended Backus-Naur Form) (Wirth,
4.5.1 Initial Rules
The new CORE-SL structure includes the initial
rules, which are defined by the models: phonetic,
morphological, syntactic, and discursive The
phonological model was the initial one proposed in
(Antunes, 2015). Thus, the morphological model
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
was defined as the ”lex-libras”, which in turn is a
”morpheme”. MORPHEM: defines that this can be
inflection, no inflection or even a NME. The rule
also describes that the three types of morphemes
present suspension, categories, identification e case;
NO INFLECTION: defined by a sign (phonological
model), the syntactic and discursive components are
under construction. Figure 3 presents this proposal.
Figure 3: Initial rules.
4.5.2 Rules for Case and Inflected Verbs
The INFLECTION define that signs that inflect and
are formed by the concatenation of a lexeme and
a grammeme; the CATEGORIES concept defines
the grammatical and morphological categories, and
the SUSPENSION concept defines the relationship
between the hands (Antunes, 2011) and sequence of a
sign (when it is a compound sign). Figure 4 presents
these concepts and attributes.
Figure 4: Rules for inflection e categories.
4.5.3 Grammatical Rules for the Formation of
Lexeme and Its Case
The following attributes have been represented for
determining the grammatical categories that specify
the syntactic-semantic-discursive rules: case, lexeme
and grammar CASE: defined by two syntactic
and semantic rules, which reflect the thematic
roles of a verbal network scheme (Felipe, 1998a),
(Felipe, 1998b). LEXEME: defines that it is
formed by the concatenation of four BSL parameters;
MOVEMENT: defines that the movements can be
local or displacement; GRAMMAR: defines the
grammemes of verbal inflection addressed in this
research. Figure 5 presents this set of rules.
Figure 5: Case Rules, lexeme and grammeme.
4.5.4 Grammeme Rules for Verbs with Person of
Speech Inflection
The PS GRAMMAR rules define that the inflection
for the person of speech, and it is carried
out through directionality, proximity, and person:
DIRECTIONALITY: defines the marking of the
people involved in the speech (Felipe, 2006);
PROXIMITY: defines the beginning and ending of the
movement, and both present the same terminal values;
PERSON: formed by the values which refer to people
involved in the speech. Figure 6 presents these rules.
Figure 6: Inflection rules for person of speech.
4.5.5 Grammeme Rules for Verbs with Gender
The rules: GENDER GRAMMAR, defined by
handshape and classifier; HANDSHAPE, defined
by ”right-hand “and” left-hand”, present values
Lex-Libras: Morphosyntactic Model of the Brazilian Sign Language to Support a Context-based Machine Translation Process
which relate to hand configurations
that form
the classifiers, and CLASSIFIER, which defines the
classifier types. Figure 7 presents these rules.
Figure 7: Inflection rules for gender.
4.5.6 Grammeme Rules for Verbs with Locative
The LOC AGREE rules, defined by localization
and action, formalize the location referring to
the place in relation to the body where the
signal will be articulated. The action represents
the performances of a speaker when signaling;
PROXIMITY defines the distance in relation to
the speaker’s body where the grammeme will be
to the exact place with marked points in relation
to the speaker’s body; CENTRAL LOCALIZATION
defines the place in relation to the parts of
the speaker’s body where the grammeme will be
articulated. Figure 8 presents this set of rules.
Figure 8: Inflection rules for the locative.
4.5.7 Rules for Morphological Functions of
According (Felipe, 2013) NME can have
phonological, morphological, syntactic, semantic,
The values of the hand settings were consulted
(Felipe, 2002) BSL Dictionary Version 2.0. Available
at: http://www.ines.gov.br/dicionario-delibras/main site /
and discursive functions. To represent these
expressions, there are the following categories:
TYPE, which defines the morphological functions
of the NME, and IDENTIFICATION, which defines
how the morphemes will be written. Figure 9 shows
these categories.
Figure 9: Rules for NME.
The Lex-Libras just presented is part of the CORE-SL
and one of the bases for the HCI-SL architecture. Its
main contribution consists of the specification of a
model that can be computerized for an automatic or
semi-automatic translation. Such a model aims to
serve as input to the “intelligent avatar”. In addition,
it also works as an input to the next linguistic level,
the syntactic one.
Lex-Libras has been theoretically validated by
a Libras computational linguist, as it includes
the Morphology Theory of BSL, thus adding
linguistic knowledge to the construction of the
model. Lex-Libras will allow to computational
vision (recognition and transcription of signals for the
morphosyntactic structure), processing of natural sign
languages, and synthesis (3D avatars).
Additionally, Lex-Libras proved the hypothesis
raised by (Antunes, 2015), that it is possible to
describe the morphosyntactic level by means of the
CORE-SL gestural-spatial parameters. In this way,
Lex-Libras is considered to be partially integrated
into the CORE-SL formal model. Never-the-less,
the full integration will only take place when the
remaining Libras linguistic levels are specified and an
experimental framework enables the intelligent avatar
together with the overall testing.
To synthesize the proposal, an abstraction of
the use of the Lex-Libras in HCI-SL will involve
(1) The user, who interacts with the Architecture
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems
through the application layer, selects a data unit,
such as the sign generator module (Avatar); (2) The
user enters a particular sign; (3) The CORE-SL
receives the request and recognizes that this signal
is composed by an inflection process; (4) The
CORE-SL, then, forwards the request to Lex-Libras,
which is responsible for the morphological module,
and looks for the sign; (5) The Lex-Libras supplies
the adequate descriptions in JSON format for the
application use, as shown in 10 below.
Intelligent Avatar Synthesis
Virtual Learning Enviroment
Libras Processing
Figure 10: Abstract of a usage of Lex-Libras in HCI-SL.
The phonological model of the CORE-SL is,
hypothetically, universal regarding the SL, although
they may vary regarding hand configurations and
the specificity of the phonological rules for each
language, as it also happens with the oral-auditory
languages. These languages are described from
the international phonetic alphabet, although each
language uses specific phonemes within these
Regarding the hypothesis of the CORE-SL formal
model being universal within sign languages, this
paper has brought, as an additional contribution, the
methodological steps to build the morphosyntactic
rules for other sign languages. The research
described in this paper is innovative as a multi, inter,
and transdisciplinary work. This dialogue among
disciplines happens in the evaluation of related results
and in the construction of the rules of the different but
interconnected linguistic levels.
The research reported in this article took a
bottom-up approach, starting from the phonological
description to the morphosyntactic one, through the
Lex-Libras, which is a formalism to generate lexical
items and organize them in the form of lexemes and
grammemes. This was based on the 5 phonological
parameters and used the phonological model as an
input to the morphological level.
The construction of the formal model with all the
internal and external properties and initiated by the
phonological level allowed the cataloging of signs in
a textual standard representation. Thus, it provides for
the extension of the rules of the successive linguistic
levels, where the previous level functions as input to
the immediately subsequent level (Antunes, 2011),
(Antunes et al., 2015), (Silva, 2020).
The architecture of hypothesis, HCI-SL, considers
the future generation of an “intelligent avatar” to
interpret the registered signal. Deaf communities
have an intuitive knowledge of their language. As
in any translation endeavor, in the construction of
rules of equivalence, it is necessary to count on
the contribution and the perspective of experts from
specific areas. It is also crucial to consider the
participation of deaf bilinguals who master such
specific areas of knowledge. Information on sentence
generation problems by existing translators is already
available (Silva, 2020).
The referred knowledge was produced by the
perception of their own deaf communities that,
yet, cannot determine which rule and at which
specific level the problems occur. Nonetheless,
deaf communities will benefit from this work as
an input to future participatory workshops. On
the other hand, listeners who use the translate
engines, when signaling the searched sentences, will
be committing the same grammatical errors of the
avatars, since they are unaware of the grammatical
rules of BSL. In this way, Lex-Libras can add
more quality and assertiveness to translation in the
future and, thus, provide better accessibility and
communicability resources to the Deafs by providing
access to information and knowledge in BSL.
In this research, similar works that focus on
other SL of other countries were also raised. Since
they have their own grammatical-discursive rules, we
did not carry any analysis of generated sentences
in different SL. However, even in different SL, the
HCI-SL Architecture can still be used, modifying the
rules of its levels that, in this paper, are specific to
BSL. The extension of this work based on a corpus
, like those presented in (Iatskiu, 2019), (Silva, 2020),
built in a formal computerized way, is part of future
work, as is the module of generation by an avatar.
Additionally, the results of the research reported
here will be appropriate by games and educational
applications already developed (Canteri et al., 2019)
For a more in-depth study involving PLN, it is
necessary to use a linguistic corpus, which is the set of
written texts and oral records in a given language and that
serves as a basis for analysis. This approach is currently
known as Corpus Linguistics, since the interpretation of the
linguistic phenomenon was based on the observation of data
from corpus and not on intuitions of the analyst (Felipe,
Lex-Libras: Morphosyntactic Model of the Brazilian Sign Language to Support a Context-based Machine Translation Process
and under development for deaf communities.
Amaral, W. (2012). Sistema de Transcric¸
ao da L
Brasileira de Sinais Voltado a Produc¸
ao de Conte
Sinalizado por Avatares 3D. PhD thesis, Universidade
Estadual de Campinas. Faculdade de Engenharia
etrica e de Computac¸
ao, Campinas, SP.
Antunes, D. (2011). Um modelo de descricao
computacional da fonologia da lingua de sinais
brasileira. Master’s thesis, Programa de Pos-Graduc¸
em Inform
atica, Curitiba- PR.
Antunes, D. (2015). Proposta de um Modelo
Computacional para Representac¸
ao de Sinais
em uma Arquitetura de Servic¸os HCI-SL para L
de Sinais. PhD thesis, Programa de Pos-Graduac¸
em Inform
atica do Setor de Ciencias Exatas da
Universidade Federal do Paran
a, Curitiba - PR.
Antunes, D. R., Guedes, A. L. P., and Garc
ıa, L. S.
(2015). A context-based collaborative framework
to build sign language databases by real users. In
Antona, M. and Stephanidis, C., editors, Universal
Access in Human-Computer Interaction. Access
to Interaction, pages 327–338, Cham. Springer
International Publishing.
ujo, T. (2012). Uma Soluc¸
ao para Gerac¸
atica de Trilhas em L
ıngua Brasileira de Sinais
em Conte
udos Multim
ıdia. PhD thesis, Tese de
doutorado. Universidade Federal do Rio Grande do
Norte, Natal, Brasil, Rio Grande do Norte, Natal.
Canteri, R., S
anchez, L., Felipe, T., Galv
ao, L., and
Antunes, D. (2019). Conceptual framework to support
a web authoring tool of educational games for deaf
De Martino, M., Silva, I.and Bolognini, C., Costa, P.,
Kumada, K., Coradine, L., Brito, P., do Amaral, W.,
Benetti, A., Poeta, E., Angare, L., Ferreira, M., and De
Conti, D. (2017). Signing avatars: making education
more inclusive. Universal Access in the Information
Society, 16(3):793–808.
Felipe, T. (1998a). A Relac¸
ao Sint
antica dos
Verbos L
ıngua Brasileira de Sinais (LIBRAS. Vol I.
PhD thesis, Programa de P
ao da Faculdade
de Letras da Universidade Federal do Rio de Janeiro,
Rio de Janeiro - RJ.
Felipe, T. (1998b). A Relac¸
ao Sint
antica dos
Verbos L
ıngua Brasileira de Sinais (LIBRAS). Vol II.
PhD thesis, Programa de P
ao da Faculdade
de Letras da Universidade Federal do Rio de Janeiro,
Rio de Janeiro - RJ.
Felipe, T. (2002). Sistema de flex
ao verbal na
libras: Os classificadores enquanto marcadores de
ao de g
enero. Anais do Congresso Surdez e
os-Modernidade: Novos rumos para a educac¸
brasileira, pages 37–58.
Felipe, T. (2003). Projeto Tradutor Libras. Technical
report, FENEIS - Federac¸
ao Nacional de Educac¸
ao e
ao dos Surdos.
Felipe, T. (2006). O processo de formac¸
ao de palavras na
libras. ETD – Educac¸
ao Tem
atica Digital, Campinas,,
pages 199–216.
Felipe, T. (2007). Descric¸
ao da l
ıngua de sinais: desafios
oricos e pr
aticos 1.
Felipe, T. (2013). O discurso verbo-visual na l
brasileira de sinais Libras / The verbal-visual
discourse in Brazilian Sign Language Libras.
Bakhtiniana. Revista de Estudos do Discurso,
url=http://ref.scielo.org/cjh56s, 8(2):67–89.
Felipe, T. (2014). Banco de Dados e Sistemas de
ao para as L
ınguas de Sinais, Instrumentos
ısticos: usos e atualizac¸
oes, Orgs: Balbas and
A. Caldas, pages 155–188. Cartolina Editora.
Ferreira, M. and Garc
ıa, L. (2018). Requirements for
avatar in virtual environment of support learning in
the literacy of deaf people in portuguese mediated
by libras. In Proceedings of the 17th Brazilian
Symposium on Human Factors in Computing Systems,
IHC 2018, New York, NY, USA. Association for
Computing Machinery.
Finau, R. (2004). Os sinais de tempo e aspecto na Libras.
PhD thesis, Programa de P
ao em Letras da
Universidade Federal do Paran
a, Curitiba-PR.
Garcia, L., Guimar
aes, C., Antunes, D., and Fernandes,
S. (2013). HCI architecture for deaf communities
cultural inclusion and citizenship. ICEIS 2013 -
Proceedings of the 15th International Conference
on Enterprise Information Systems, 3(October
Iatskiu, C. (2019). Core-SL-SW-generator: gerador
atico da escrita da libras a partir de um
modelo de especificac¸
ao formal dos sinais. PhD
thesis, Programa de Pos-Graduac¸
ao em Inform
do Setor de Ciencias Exatas da Universidade Federal
do Paran
a, Curitiba - PR.
Lima, M. (2015). Traduc¸
ao autom
atica com adequac¸
antica para libras. Master’s thesis,
Centro de Inform
atica da Universidade Federal da
ıba, Jo
ao Pessoa - PB.
Paiva, F. (2019). Faculdade de Engenharia El
etrica e
de Computac¸
ao Traduc¸
ao Autom
atica de Portugu
Brasileiro para Libras e An
alise de Processos de
ao Traduc¸
ao Autom
atica de Portugu
Brasileiro para Libras e An
alise de Processos de
Silva, A. (2020). Descric¸
ao das unidades morfossint
da libras atrav
es dos par
ametros fonol
ogicos para
o processo de traduc¸
ao autom
atica por um avatar.
Master’s thesis, Programa de P
os Graduac¸
ao em
atica - Universidade Federal do Paran
Curitiba - PR.
Stokoe, W. (1960). Sign Language Structure. An Outline
of the Visual Communication Systems of the American
Deaf. SpringLinstok.
Wirth, N. (1977). What can we do about the
unnecessary diversity of notation for syntactic
definitions? Communications of the ACM,
ICEIS 2021 - 23rd International Conference on Enterprise Information Systems