Discovering Emotions through the Building of a Linguistic Resource

Manuela Angioni and Franco Tuveri

Center for Research and Scientific Studies in Sardinia, Bld. 1, Piscina Manna, Pula (CA), Italy

Keywords: Linguistic Resources, Natural Language Processing, Sentiment Analysis.

Abstract: Specific linguistic resources, syntactically annotated and distinctive for each language, related to the

affective sphere are important in discovering terms or phrases associated with emotions in order to detect

expressed emotions. The paper proposes the initial version of a linguistic resource for the Italian language,

mapped on WordNet, where each concept, whose meaning falls into the sphere of emotions, is enriched by a

category, allowing to better specify the type of emotion expressed by the term, and by a polarity value,

whether the emotion is positive or negative. The resource is based on the model of emotions proposed by

Robert Plutchik and has been developed, within a national project of Work-School Alternation, in

collaboration with some high school students. The work has a twofold value. On one hand, the development

of a linguistic resource, on the other the educational and didactic aspect of students’ involvement. Working

on the analysis of literary texts with the task of elaborating and defining the emotions described, the

students, assisted by their teachers and two researchers, had to face with their feelings and talk more freely

about their affective states, recognizing the emotions and giving them a name.

1 INTRODUCTION

Emotions constitute a fundamental aspect of people's

lives. All the things people do or say reflect their

emotions somehow and express it in all the

communication ways. The automatic identification

of emotions in a collection of textual data is

becoming increasingly important, in order to be able

to understand in more detail the mood and the

sentiments and to better interpret the human

experience and the type of communication expressed

in the texts.

The building of linguistic resources is part of

Computational Linguistics, the discipline that deals

with the study of natural language by means of

automatic tools, through the definition of algorithms

that allow the automatic analysis of texts and the

extraction of the meaning expressed in them

The analysis of the texts involves several phases,

including syntactic and semantic analysis of the

sentences in the text. A further fundamental step to

detect emotions in a text is to identify terms or

phrases dealing with affective and emotional sphere.

To this end, the availability of linguistic resources,

syntactically annotated and specific for the language

used in the text, is necessary.

The paper proposes the creation of a linguistic

resource for the Italian language named

PlutchikWN_ita, based on the model of emotions

proposed by Robert Plutchik (Plutchik, 1980),

whose terms are related to the emotions and mapped

on WordNet (Miller, 1995). Each concept is

enriched by a category, that allows to better specify

the type of emotion expressed by the term, and by a

polarity value. The resource is built on the basis of

the experience of the daily discovery of the emotions

of some adolescent students and the impact of

emotions on their lives.

The work therefore has a twofold value.

On one hand, the development of a linguistic

resource that allows a deeper analysis of textual

resources, through the recognition of terms related to

the emotional sphere and the type of emotion

expressed.

On the other hand, the educational and didactic

aspect of students’ involvement. In the school

environment it is in fact essential to promote a

culture in which students can learn to discover their

sentiments and moods, and teachers and students

together can freely talk about feelings. Speaking

about emotions, it is important to recognize the

feelings that the students have, so that they are able

to give them a name. It is thus important to learn

students to identify their different moods and

feelings educating them in recognize, manage, and

express their emotions.

Angioni, M. and Tuveri, F.

Discovering Emotions through the Building of a Linguistic Resource.

DOI: 10.5220/0008352803510357

In Proceedings of the 15th International Conference on Web Information Systems and Technologies (WEBIST 2019), pages 351-357

ISBN: 978-989-758-386-5

351

To this aim, in agreement with the teachers,

some students were involved in the analysis of some

literary texts, forming part of their course of study,

with the task of elaborating and defining the feelings

and the emotions described and aroused by the

reading of the texts. From the dialogue with the

teachers and the class emerged in fact the common

difficulty among students of similar age in

recognizing and naming their emotions, as well as in

speaking openly about them.

It was therefore decided to include the creation

of the linguistic resource in a broader context,

combining the strongly multidisciplinary aspect

characteristic of this activity that requires a

heterogeneous workgroup and a constant dialogue

between professionals with very different skills.

The building of such linguistic resource

combines in fact humanistic skills, to address

aspects closer to Linguistics or Psychology, with

others coming from Communication and

Information Technologies (ICT). This made it

possible to establish a dialogue on several fronts

with the students, involving them in a

multidisciplinary project.

The paper describes also the experience

conducted in collaboration with the third year

students of the “Liceo Classico, Linguistico e delle

Scienze Umane B. R. Motzo” of Quartu Sant’Elena,

Cagliari. The activities are part of the CRS4

initiatives within a national project of Work-School

Alternation, which aims to promote the training of

students in the field of digital technologies and

scientific research.

The activities took place with a presentation of

the work in front of the whole class through the

involvement of all the students and the introduction

of the topics concerned. A smaller group of three

students worked on the realization of the linguistic

resource through the skills of two CRS4 researchers,

who provided the assistance and the training

necessary to carry out the assigned task.

The remainder of the paper is organized as

follows: Section 1 introduces the resource and the

motivation behind. Section 2 describes the related

works. Section 3 presents the building of the

resource, with an overview of the Plutchik’s Wheel

of Emotions, and describes the interface used.

Section 4 describes the current state of

PlutchikWN_ita, while Section 5 presents the format

of the resource. Finally, in Section 6 final remarks

and future works directions are presented.

2 RELATED WORK

Several resources about emotions exist for the

English language, such as NRC Emotion Lexicon

(Mohammad and Turney, 2010), composed by

frequent English terms annotated through Amazon

Mechanical Turk with respect to eight emotions (e.g.

joy, sadness, trust) and positive or negative

sentiment. Another resource, WordNet-Affect

(Valitutti et al., 2004), has been developed starting

from WordNet, assigning one or more affective

labels (a-labels) to a subset of synsets representing

affective concepts that contribute to precise the

affective meaning. In WordNet-Affect, Ekman’s six

basic emotions (anger, disgust, fear, joy, sadness,

surprise) have been used.

Other affective lexicons are available but they

often give a polarity evaluation of terms without a

reference to an emotional category, such as

SentiWordNet (Baccianella et al., 2010) devised for

supporting sentiment classification and opinion

mining. SentiWordNet is one of the publicly

available lexical resources, that extends WordNet

thanks to a semi-automatic acquisition of the

polarity of WordNet terms, evaluating each synset

according to positive, negative and objective values.

In this scenario we choose to develop a new

resource based on the model of emotions proposed

by Robert Plutchik (Plutchik and Conte, 1997),

(Plutchik 2002), an American psychologist,

Professor Emeritus at the Albert Einstein College of

Medicine and Adjunct Professor at the University of

South Florida. He developed a psychoevolutionary

theory about the emotions and created a model based

on a distinction between basic and complex ones

(Kołakowska et al., 2015), introducing an emotion

classification system known as the Wheel of

Emotions.

3 BUILDING THE RESOURCE

The idea of the resource, named PlutchikWN_ita,

arises from the experience gained in the field of

Opinion Mining and of the automatic analysis of

texts through a linguistic approach in which

WordNet has been chosen as the main linguistic

resource (Tuveri and Angioni, 2012a).

We realized at that time the need to enrich

WordNet through some additional properties,

initially linked to adjectives and adverbs and later

extended to verbs, to get a deeper analysis of the

sentences and improve the semantic disambiguation

WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies

352

phase. A first resource has therefore developed

(Tuveri and Angioni, 2012b) in which the synsets

related to adjectives and adverbs are enriched with a

group of properties and with a positive, negative and

objective polarity value. The building of this first

linguistic resource, related to adjectives and adverbs,

and the analysis of the properties of the terms

already included in WordNet and identified from the

lexicon files, revealed the need to better specify the

properties of the terms inherent to sentimental and

emotional states. In fact, in both the resources there

are properties related to emotions, such as

noun.feeling and verb.emotion in WordNet or

emotion and morals/ethics in the resource we

developed.

For this reason we have built a linguistic

resource that allows to better specify the type of

emotion expressed by the term extending the

WordNet properties with the categories arising from

the Plutchik model.

Despite the fact that the resource is very limited

in number of terms and concepts, its construction

allowed to face some of the problems related to a

resource dealt with the emotional sphere and to lay

the foundations for the development of a more

complete resource.

Although there is no fully exhaustive and

universally accepted model of emotions, and looking

for a substantially simple model for the students, it

was decided to consider the model of emotions

proposed by Robert Plutchik. In fact, the model

defines a few basic emotions, making the problem of

the development of the resource manageable, and

having as strength point its capacity to simplify very

complex concepts.

Moreover, the graphical representation of the

emotions proposed by Plutchik enable to

immediately visualize the emotions, and thus easily

understand which combinations of emotions created

a resulting emotion.

3.1 Plutchik’s Wheel of Emotions

Plutchik proposes a graphic representation of the

wheel of emotions that, starting from a circle of

eight primary and fundamental emotions, develops

in the structure shown in Figure 1.

This model shows how different emotions can be

combined or mixed together, as could do a painter

with colors. According to his theory, other emotions

are a combination of the basic emotions. Plutchik's

studies have therefore highlighted 8 primary basic

emotions: joy and sadness, trust and disgust, anger

and fear, surprise and anticipation. Primary emotions

are represented through a circle, in which the closest

emotions are also the most similar and opposite

emotions are located on opposite spokes of the

wheel. So joy is opposed to sadness, anger is

opposed to fear, acceptance is opposed to boredom,

and surprise is in opposition to anticipation.

Figure 1: Plutchik's Wheel of Emotions.

Plutchik places the emotions similar to the

primary ones, but of lower intensity, on the external

part of the wheel, putting them all the more distant,

the less their intensity. Following the same principle,

similar but more intense emotions are placed within

the wheel of primary emotions. In this way, a kind

of flower is obtained whose petals can be joined

towards the bottom creating a three-dimensional

representation in a cone-shaped model.

Figure 2: Primary emotions combination or dyads.

Also the variations in the intensity of the color

on each petal correspond to the variations in

Discovering Emotions through the Building of a Linguistic Resource

353

Figure 3: Insertion of Italian terms and Pluchik’s emotions.

intensity of the emotions. The eight primary

emotions occupy the central part of the flower, with

gradually less intense emotions by proceeding from

the center towards the end of the petal. For example,

the most intense form of fear is terror, the least

intense is apprehension.

Figure 2 shows how primary emotions work in

combination. The compositions of two emotions are

called dyads. All not primary emotions are mixed or

derived, that is, they appear as combinations or

compositions of primary emotions and are identified

as primary, secondary or tertiary dyads depending

on whether adjacent emotions are combined, shifted

by one or two petals.

Primary dyads made from adjacent primary

emotions are considered “often felt”. The emotions

can then be combined in a variety of ways. Dyads

including emotions further away on the wheel,

express less felt emotions, until opposites create

conflict. For example, joy and trust could combine

to create love, joy with fear could give rise to guilt,

anticipation combined with fear could give rise to

apprehension (Taylor, 2017).

3.2 The Identification of Emotional

Terms

The building of the resource PlutchikWN_ita has

been performed by means of an ad-hoc developed

web-based application, that allows to insert one or

more terms in Italian and to specify both a polarity

value and an emotion from the list of Plutchik’s

categories, as shown in Figure 3.

The search for Italian terms, which fell within

the affective and emotional sphere, was conducted

by involving the students.

They examine some of the most famous carmina

written by Catullo and the Italian version of the “La

notte d’amore tra Ginevra e Lancillotto”, from the

12th-century Old French poem “Lancelot ou le

Chevalier à la charrette” written by Chrétien de

Troyes, highlighting the terms of interest through

reading.

The use of these terms in the texts allows the

author to communicate his mood to the reader. In

them, therefore, the expressive function prevails and

the attention focuses on emotions, thoughts and

personal experiences.

Once identified an Italian term falling within the

emotional sphere, the term in English with the

corresponding meaning is detected, making use of

several online dictionaries when necessary.

The application shows all the possible meanings

included in WordNet pertains to the emotional

sphere or to morality, starting from the identified

English term and from the relative Part of Speech

(POS). At this point, it is possible to insert in the

resource PlutchikWN_ita the Italian term and its

synonyms, corresponding to the English term and to

the definition, and to specify the Plutchik’s category

that best reflects the type of emotion expressed by

the term.

It was decided to exploit the terms in English

because the English version of WordNet is the most

WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies

354

complete both as a number of concepts and

definitions.

Each of the students had the task of

independently inserting each term in Italian and the

relative Plutchik’s emotion. Once the insertion was

completed, the terms and the categories were

validated.

When disagreement occurred in the identification

of the emotion, the students discussed the

motivations that led to the specific assignment. In

some cases a convergence of opinions has been

found, otherwise the term was not included in the

resource.

Through the analysis of the texts the students

have identified the words expressing the emotions,

relating them to the context and using them as a

starting point to investigate the real meaning of the

emotion by means of discussion and argumentation.

The recognition of the emotions of the others refined

their ability to identify themselves with the moods of

others. At the end of the work the students also

showed a greater awareness towards their emotions

and a greater attention to their inner states.

4 THE CURRENT VERSION

The concepts included in the resource are currently

very limited, 214 in total, and are divided between

names (N), adjectives (A), adverbs (R) and verbs

(V), as shown in Table 1.

It is possible to notice that the majority of the

terms are related to adjectives, followed by nouns

and verbs, while terms related to adverbs are

irrelevant, but they could be easily extracted from

the corresponding adjectives and nouns.

The resource is based on WordNet 3.1 version.

Table 1: The distribution of concepts for POS.

POS

Synsets/Terms

From the analysis of the resource emerges that

the richest emotions as number of synsets are

serenity (30), grief (20) and admiration (22), as

shown in the following Table 2.

As an example, in PlutchikWN_ita the Italian

terms like tranquillo, sereno and calmo have been

classified as positive terms under the serenity

Plutchik’s emotion (serenità in Italian) and

correspond to the English words calm, serene,

tranquil, unagitated. For the sake of completeness, in

WordNet v.3.1 the synsetID is 300531862 and the

definition: is “not agitated; without losing self-

possession; "spoke in a calm voice"; "remained calm

throughout the uproar" ”.

Table 2: Plutchik’s emotions and number of synsets.

Plutchik’s

Emotions

Syns

Plutchik’s

Emotions

Syns

serenity

aggressiveness

grief

apprehension

admiration

low_love

joy

pensiveness

interest

high_aggressiveness

sadness

remorse

trust

submission

fear

high_remorse

high_love

boredom

love

annoyance

Other examples of terms included in the resource

are the Italian adjectives sofferente, penoso e

doloroso, corresponding to the English terms

afflictive, painful, sore, that have been classified

under the Plutchik’s emotion grief (dolore in Italian)

with a negative polarity. The term is related to the

synsetID 301809309 in WordNet v.3.1 and is related

to the definition: “causing misery or pain or distress;

"it was a sore trial to him"; "the painful process of

growing up" ”.

5 THE RESOURCE FORMAT

PlutchikWN_ita is currently being developed in the

WordNet LMF format (Soria, 2009), an extension of

the ISO LMF format defined for the representation

of lexical resources. LMF allows the coding of

lexicons in natural language, and provides a

common model for the creation and use of lexical.

The WordNet LMF has been introduced in the

EU KYOTO project (Vossen et al., 2008) with the

specific purpose of providing the WordNet, set in

different languages, with a standardized

interoperability format that would allow the

exchange of coded lexico-semantic information in

each of them. The use of this format enables

WordNet to provide a format representation that

allows easier integration between resources that

share the same structure and, more importantly,

between resources with different theoretical and

implementation approaches. This fact introduced a

new kind of problems especially with the enrichment

of new synsets, not included in the English version

of WordNet but specific of the language.

Furthermore, the different wordnets can be built

using different methods and starting from different

Discovering Emotions through the Building of a Linguistic Resource

355

points: with the expand or the merge approach,

manually or semi-automatically, using pre-existing

resources or original bilingual resources available

for the translation of the original English words in

the target language.

All this led to the definition of CILI, the

Collaborative Inter-Lingual Index, (Bond et al.,

2016), based on the interlingual index ILI

(Interlingual Index) and proposed for the first time

in the EuroWordNet project (Vossen, 1998).

CILI is a single shared repository of concepts,

defined as an inter-collaborative index designed to

make possible to coordinate multiple generic

wordnet projects in the Global WordNet Grid

(GWG).

Figure 4: An example of PlutchikWN_ita in LMF format.

Figure 4 shows an example of the

PlutchikWN_ita in WordNet LMF format. Each new

wordnet has a Lexicon for each resource, which

contains a name, an ID and a language. The

language is specified by the Internet Engineering

Task Force (IETF) Best Current Practice (BCP) 47

language tag, the code to identify human languages.

Each term inserted in the linguistic resource is

defined as a Lexical Entry, and is characterized by a

lemma, by at least a meaning and possibly by any

number of Syntactic Behavior.

The Lemma is defined through its written form

and the Part Of Speech (POS), which can be a name

(n), a verb (v), an adjective (a), an adverb (r), a

phrase (like phrase, p, o sentence, s) or an unknown

pos (u).

The Sense has a variable number of sense

relations and a synset.

The Synset can have an optional definition and a

variable number of sense relations.

The Definition is given both in the WordNet

language from which it comes but also in the ILI

definition language (in English). A definition can

also have a sentence that provides an example of

using of the term.

Synset Relation specifies the type of relationship,

such as synonimy, hyperonymy, antonymy, taken

from the list of reports used by the Global WordNet

Grid and documented on the Global Wordnet

Association website

A Syntactic Behavior (verb frame) provides a

sub-category frame in plain text.

Metadata from the Dublin Core system can be

added to the Lexicon, Lexical Entry, Sense and

Synset tags.

6 CONCLUSIONS

The paper describes the creation of a new linguistic

resource, built on a set of terms mapped on

emotions, by means of the everyday life of the

discovery of emotions and their impact on the lives

of students in adolescent age. In agreement with the

teachers, some students were involved in the

analysis of some literary texts, forming part of their

course of study, with the task of elaborating and

defining the feelings and the emotions described and

aroused by the texts.

The reading of literary texts and the

identification in them of terms relating to the

emotional and affective sphere, allowed the students

to define their emotions by identifying themselves in

the lives of others and recognizing themselves in

other people. With this work, an alternative teaching

method was proposed that allowed students to put

the emotional intelligence at the center and stimulate

it, as a primary element of the evolution of their

personality.

The building of the linguistic resource, in a

broader context, combines the strongly

multidisciplinary aspect characteristic of this activity

that requires a heterogeneous workgroup and a

constant dialogue between professionals with very

different skills, with the didactic-educational aspect

that allowed the students to give voice to their

emotions.

Global WordNet Association, http://globalwordnet.org/

WEBIST 2019 - 15th International Conference on Web Information Systems and Technologies

356

Currently the resource is only at the initial stage,

but in the future we intend to extend the resource

through the involvement and the systematic

contribution of different classes of students

supported by their teachers. The resource has not

still been used in any practical application, but we

are planning to realize a comparative work using

similar resources, with the intent to provide its

validation results.

ACKNOWLEDGEMENTS

We wish to thank the three high school students

Katia Pedditzi, Ludovica Medved and Alessandra

Pani, their whole class, and their teachers and tutors

Cinzia Quattrocchi and Antonella Puggioni, from the

“Liceo Classico, Linguistico e delle Scienze Umane

B. R. Motzo” of Quartu Sant’Elena, Cagliari, to their

contribution to the development of the resource.

REFERENCES

Baccianella, S., Esuli, A., Sebastiani, F., 2010.

SentiWordNet 3.0: An Enhanced Lexical Resource for

Sentiment Analysis and Opinion Mining. In LREC

2010, 7th International Conference on Language

Resources and Evaluation, Malta, pp. 2200-2204.

Bond, F., Vossen, P., McCrae, J., Fellbaum, C., 2016.

CILI: the Collaborative Interlingual Index.

Proceedings of the 8th Global WordNet Conference

2016, in Bucharest, Romania, January 27-30.

Global WordNet Association Homepage,

http://globalwordnet.org/.

Kołakowska, A., Landowska, A., Szwoch, M., Szwoch,

W., Wrobel, M.R., 2015. Modeling emotions for

affect-aware applications. In book: Information

Systems Development and Application, Chapter: 4.

Publisher, Faculty of Management University of

Gdańsk. Editors: Stanislaw Wrzycza.

Miller, G. A., 1998. WordNet: A Lexical Database for

English. Communications of the ACM Vol. 38, No.

11: 39-41.

Mohammad, S.M., Turney, P.D., 2010. Emotions evoked

by common words and phrases: Using mechanical turk

to create an emotion lexicon. In: Proceedings of

NAACL 2010 Workshop on Computational

Approaches to Analysis and Generation of Emotion in

Text. ACL.

Plutchik, R., 1980. Emotion: Theory, research, and

experience: Vol. 1. Theories of emotion, 1, New York:

Academic.

Plutchik, R., H. R. Conte, 1997. Circumplex Models of

Personality and Emotions. Washington, DC:

American Psychological Association.

Plutchik, R., 2002. Emotions and Life: Perspectives from

Psychology, Biology, and Evolution. Washington, DC:

American Psychological Association.

Taylor, M. J., 2017. Classroom Emotions.

https://www.tetraanalytix.com/classroom-emotions/.

Tuveri, F., Angioni, M., 2012a. A Linguistic Approach to

Feature Extraction Based on a Lexical Database of the

Properties of Adjectives and Adverbs. Proceedings of

the 6th International Global Wordnet Conference

Tribun EU, pages 365-370, Global WordNet

Association. ISBN: 978-80-263-0244-5.

Tuveri, F., Angioni, M., 2012a. Definition of a Linguistic

Resource for Opinion Mining. Proceedings of the 9th

International Workshop on Natural Language

Processing and Cognitive Science –NLPCS -

SciTePress -june 2012. ISBN: 978-989-8565-16-7.

Soria, C., Monachini, M., Vossen, P., 2009. Wordnet-

LMF: Fleshing out a Standardized Format for Wordnet

Interoperability. Proceedings of the 2009 international

workshop on Intercultural collaboration. ACM.

Valitutti, A., Strapparava, C., Stock, O., 2004. Developing

affective lexical ressources. Psychnology: 2-2004.

Vossen, P. (ed), 1998. EuroWordNet: A Multilingual

Database with Lexical Semantic Networks. Kluwer

Academic Publishers, Dordrecht.

Vossen, P., E. Agirre, N. Calzolari, C. Fellbaum, S. Hsieh,

C. Huang, H. Isahara, K. Kanzaki, A. Marchetti, M.

Monachini, F. Neri, R. Raffaelli, G. Rigau, M. Tescon,

J. van Gent, 2008. KYOTO: A System for Mining,

Structuring, and Distributing Knowledge Across

Languages and Cultures. Proceedings of the Fourth

International Global WordNet Conference - GWC

2008. Szeged, Hungary, January 22-25.

Discovering Emotions through the Building of a Linguistic Resource

357