A Formal Model to Support Discourse Semantic Landscape Analysis

Isabelle Linden

, Bruno Dumas

and Anne Wallemacq

Namur Digital Institute (NADI), University of Namur, Belgium

Keywords:

Discourse Analysis, Semantic Landscape, Text Formal Model.

Abstract:

Discourse Analysis is a broadly spread methodology in human and social sciences. The Evoq Software (Clar-

inval et al., 2018) has been developed to offer advanced support for a deep semantic analysis of discourse by

providing intermediary transposition tools that allow the exploration of the semantic landscape underlying a

discourse from multiple angles. This paper presents the formal knowledge model to support these functionali-

ties development and ensure a strong coherence between multiple visualisations seen as so many intermediary

transpositions of the same object.

1 INTRODUCTION

Discourse analysis is an essential practice at the heart

of the activities of many researchers in the humani-

ties and social sciences. Unfortunately, few software

packages offer advanced features for qualitative anal-

ysis. Most of the tools available to humanities re-

searchers are limited to quantitative functions. The

qualitative approach is often neglected or limited to

text manipulation, analysis and dictionary functions

(Lejeune, 2010; Lejeune, 2021).

The EFFaTA-MeM (Evocative Framework for

Text-Analysis - Mediality Model) research project

and the Evoq software have the essential ambition of

offering advanced support for a deep semantic anal-

ysis of discourse by providing intermediary transpo-

sition tools that allow the exploration of the semantic

landscape underlying a discourse from multiple an-

gles (Linden et al., 2020).

To ensure the strong coherence of the intermedi-

ary transposition, a key question is addressed in this

paper: How to formalise the semantical elements of a

discourse to support the extraction, the visualisation

and the analysis of its semantic landscape?

To answer this research question, the paper pro-

ceeds as follows. Section 1 presents the theorical

background used to support our text analysis ap-

proach, namely the structural analysis and introduces

the notion of semantic landscape. Then, Section 3

proposes a basic formal model integrating the key ele-

https://orcid.org/0000-0001-8034-1857

https://orcid.org/0000-0001-5302-4303

https://orcid.org/0000-0001-9822-4966

ment of this theory. This model is extended in Section

4 so as to integrate the elements useful for intermedi-

ary transposition. Finally, Section 5 present the trans-

position integrated in the Evoq Software and explains

how they are supported by the formal model.

A short analysis example drawn on text presenting

the Transition Network (Transition Network, 2021)

serves as running example all along the paper.

2 STRUCTURAL ANALYSIS AND

SEMANTIC LANDSCAPE

One of the essential goal of the EFFaTA-MeM project

is to conceive ways of interacting and visualising texts

that foster new insights in their analysis. This means

that we investigate (i) the mediality of texts and of

graphical and pictorial representations (Kucher and

Kerren, 2015; Gibson et al., 2013), (ii) the intermedial

transposition from the textual semiotic system to a

pictorial semiotic (Rajewsky, 2002; Rajewsky, 2005;

Wolf, 1999; Ellestr

om, 2010; Ellestr

om, 2014), and

(iii) the meaning enrichment opportunities offered by

this transposition.

The theoretical background used to develop the

mediality studied in this research is based on a

deep reformulation of the post-structuralist principles.

From a mediality point of view, post-structuralism is

interesting because it fundamentally questions and re-

frames the linearity of the text. Post-structuralism

opens a very interesting conception of language of

which we present here some essential features for our

work.

118

Linden, I., Dumas, B. and Wallemacq, A.

A Formal Model to Support Discourse Semantic Landscape Analysis.

DOI: 10.5220/0011298300003269

In Proceedings of the 11th International Conference on Data Science, Technology and Applications (DATA 2022), pages 118-126

ISBN: 978-989-758-583-8; ISSN: 2184-285X

2.1 A Synchronous View on the Text

While the text is commonly considered as a linear

continuous deployment from a beginning to an end,

post-structuralism considers it as a ﬁeld of forces

given synchronously. It means that the beginning and

the end of the text are considered in the same way and

with the same status, just as everything inbetween.

2.2 A Relational Semantics

In this global view on the text, rather than phrases or

words, basic units are couples of words in opposition,

such as White/Black.

Moreover, each word in the opposition is wrapped

up in a set of associative relations or simple evoca-

tions which may be explicitly formulated in the text

or implicitly supposed by the culture or the context.

So for example, in the White/Black opposition White

is associated pure and good while Black brings with

it implicitly dirty and bad. The meaning of words is

therefore larger than the simple denotation. White is

far more than a simple colour. Opposed to black it

conveys the semantic universe of purity, angel, par-

adise, untouched.

Let’s now consider another text in which white ap-

pears in tension with red. Red introduces associations

with life, warmth,... to love or anger. In tension with

this red, white then becomes the bearer of death, cold,

impassive,... radically different from its meaning

in the previously considered opposition white/black.

These considerations highlight the deeply relational

nature of word semantics in the structuralist perspec-

tive, which is no longer intrinsic to the word under

consideration but depends on its interweaving in the

network of words in the text

2.3 The Semantic Fields

The concept of semantic ﬁeld has been proposed to

cover this system of oppositions and associations. Se-

mantic ﬁelds are not always explicitly revealed by the

words in the text, they are commonly assumed, re-

sulting from the culture or the speciﬁc context. In the

structuralist approach, this distinction is referred to as

the distinction between the level of language and the

level of discourse. It implies that an author (writer or

speaker) never has a complete mastery of the mean-

ing of her words. Indeed, her discourse is received by

the audience integrating these semantic ﬁelds that are

collectively produced and taken for granted. The se-

mantic ﬁeld has thus to be considered as the semantic

surrounding the text.

Going even further, Derrida (Derrida, 1967) sug-

gests that these semantic ﬁelds are not a quiet equi-

librium but always in power tension according to the

idea that there are dominant relationships between

competing semantic ﬁelds.

2.4 Structural Analysis

By studying the semantic ﬁelds which are underlying

the explicit and conscious discourse, structural analy-

sis aims to reveal the balance of power in deﬁning the

dominant worldview underlying a text.

3 BASIC CONCEPTUAL MODEL

In the post-structuralist approach as presented in the

previous section 2, a text becomes much more than a

linear sequence of words. It involves a domain knowl-

edge, implicit representations and tensions which are

captured by relations. This section presents the basic

formal model developed to capture this approach and

formalise a notion of Analysis project which extends

the notion of text according to the principles of struc-

tural analysis.

This formal model plays two key roles in the over-

all framework of the research:

• ﬁrst, it is itself a transposition on which artiﬁcial

intelligence techniques can be applied.

• then, it can be used to ensure the coherence be-

tween the multiple proposed intermedial transpo-

sitions,

After the introduction of the underlying Intuitions

subsection 3.1, subsection 3.2 introduces the notation

of the basic model and subsection 3.3 formalises the

model transformations reﬂecting the steps of an anal-

ysis process.

3.1 Intuitions

The synchronous nature of the approach adopted al-

lows the consideration of the analysis as a single ob-

ject without any temporal aspect. As the main focus is

on relationships, a key element of our model consists

of a set of relationships between objects that, for sake

of simplicity we call words. Actually, there are three

different types of words depending of the object of in-

terest for the analyst. The ﬁrst ones are single literal

expression (as e.g. a brand name or ”Transition” in

the context of our illustration). The second ones are

equivalence classes on the natural language dictionary

for which a root represent itself and all its derivative.

A Formal Model to Support Discourse Semantic Landscape Analysis

119

And, the last ones are concepts represented by expres-

sion involving several words (as e.g. ”collective intel-

ligence”). In our model, we denote by Word the set

of all these possible values.

Relations between these words can be of two

different natures: associations denoting a proximity

(whatever its nature), and oppositions that materialise

the tensions of the semantic ﬁeld. The description of

the approach does not suggest any orientation of the

relations, consequently they are modelled by symmet-

rical relations.

3.2 Notations

The basic elements of our model are build from three

sets: Text which denotes the texts that can be anal-

ysed, Word introduced above, and BasicRel the set

of relations, formalised as (Word ×Word ×Boolean).

With these notations, the basic model of an analysis is

a triple

(t, wl, rl) ∈ Text ×P(Word) × P(BasicRel)

where

• t is the analysed text,

• wl is the set of the words of interest,

• rl is the set of the relations,

• all the word appearing in rl are in wl

• (w

, w

, b) ∈ rl iff (w

, w

, b) ∈ rl.

Note that this model integrates associations and

oppositions in one single set. They can be retrieved

as {(w

, w

) : (w

, w

, b)} with b = 0, 1 respectively.

3.3 Operations

With this notation, drawing an analysis consists in be-

ginning with (t,

0) and integrating information step

by step by either integrating a new word w or a new

relation r = (w

, w

, b) in the analysis (t, wl, rl). The

addition of a word is formalised by the following op-

eration.

addWord : ((t, wl, rl), w) → (t, wl ∪ {w}, rl)

The addition of a relation is a partial function only

deﬁned provided none of (w

, w

, 0) or (w

, w

, 1) is

in rl:

addRel : ((t, wl, rl), (w

, w

, b))

* (t, wl ∪ {w

, w

}, rl ∪ {(w

, w

, b), (w

, w

, b)}

but not mandatory in t

The operations set involves also possibility for

correction in an analysis, namely, removing a term or

a relation.

removeWord : ((t, wl, rl), w)

→ (t, wl \ {w}, rl \ {(w

, w

, b) : w

= w ∨ w

= w}

removeRel : ((t, wl, rl), (w

, w

, b))

→ (t, wl, rl \ {(w

, w

, b), (w

, w

, b)})

Note that to preserve the constraints on the inclu-

sion of words appearing in the relations in the set of

words, removing a term implies the withdrawal of all

the relations in which it is involved.

4 INTERMEDIAL CONCEPTUAL

MODEL

In an unpublished study of the ﬁrst paragraph of

a description of the Transition (Transition Network,

2021), the Words dictionary made of 21 words (Head,

Heart, Hands, ﬁnd, best information, evidence avail-

able, collective intelligence, compassion, value, pay-

ing attention, emotional, psychological, relational, so-

cial aspect, tangible reality, practical projects, build,

new healthy economy, good intentions, charity, old

economy) and 26 symetrical relations between them.

Intermediate transposition aims to present the

same information in different media in a way that fa-

cilitates its appropriation by the analyst while stim-

ulating reﬂection. In section 5, we explore several

visual transpositions. Preliminary, this section de-

scribes the enrichment of the conceptual model to

take into account visual aspects of the transposition

regardless the speciﬁc visual format.

After the introduction of the Analysis Project

model in subsection 4.1, subsection 4.2 enhances the

enriched views on the project which can be derived

from the model. Then, section 4.3 formalises the

model transformations and their use in the process of

drawing an analysis.

4.1 The Knowledge Model

Integrating the elements useful for conducting a struc-

tural analysis as described in the section 2, the text-

enriched model, which we call Analysis Project (or

AP) consists of a Text, FieldWords, FieldRelations,

and mappings to a representation domains deﬁned by

functions of DomMap. Let’s ﬁrst introduce these sets.

Notation 1. Let

• FieldWord = Word × Colour denote the set of

pairs of words associated with a colour,

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

120

• FieldRelation = Word ×Word × Boolean denote

the set of pairs of words associated with a boolean

denoting if the relation is a association or an op-

position

• DomMap = FieldWord * Coordinate denote the

set of the partial functions deﬁned on (a sub-

set of) Fieldword into a visualisation domain

Coordinate.

Deﬁnition 1. Formally, the minimal deﬁnition

of a analysis project can be given by a tu-

ple (t, wd, rd, ml) in AnalysisProject = Text ×

P(FieldWord) × P(FieldRelation) × P(DomMap)

where

• t ∈ Text, is the text object of the analysis,

• wd ∈ P(FieldWord) is the set of purposeful words

associated with colours. wd is also called the

Words dictionary,

• rd ∈ P(FieldRelation) is a list of pairs of Words

appearing in wd with a boolean denoting if the re-

lation reﬂects an evocation or an opposition (also

called association and dissociation). rd is also

called the Relations dictionary

• ml ∈ P(DomMap) is a list of partial function de-

ﬁned on (a subset of) the words in wl into (possi-

bly different) visualisation domains.

A few notations are associated with this model.

Notation 2. Given a words dictionary wd, and a spe-

ciﬁc FieldWord w, we denote by

• wd

, the i

element of the list,

• w.word, the word item of the FieldWord pair.

• w.colour, the colour item of the FieldWord pair.

Similarly, given a relations dictionary rd, and a spe-

ciﬁc FieldRelation r we denote by

• rd

, the i

element of the list,

• r.w1, the ﬁrst word item of r,

• r.w2, the second word item of r,

• r.rel, the boolean denoting the type (association

or opposition) of r.

Beside the project analysis itself, the analysis

drawing process, as well as its presentation integrates

knowledge (k ∈ Knowledge) which basically consists

in structured knowledge as dictionaries, lemmatisers,

synonyms dictionaries, antonyms dictionary and for-

malised domain knowledge as well as explicit and im-

plicit expert knowledge.

For sake of simplicity, we call here ”word” the the ba-

sic unit used in the analysis. In a linguistic perspective, it

could be more properly called ”lexical item”. Indeed, ac-

cording to the analyst’s object of interest it can be either a

speciﬁc word, a lemma or a ﬁxed set of words

4.2 Enriched Views of an Analysis

Project

Based on the minimal representation introduced in the

previous section, enriched concepts are built that in-

tegrate various element of the Analysis Project Field

tuple. A few notations are introduce to support their

formal deﬁnition.

Deﬁnition 2. Given a word w, exploiting the linguis-

tic knowledge formalised by a dictionary in k, we call

deriv(w) the set including all its lexical derivation if

w is a lemma, the singleton involving only w if w is a

derived word or a compound expression.

In the following, we extend the seminal notation

∑

i=1

to the Append function on list

Notation 3. We denote by

Append

i=1

description(...i...)

the list composed of the n elements obtained by in-

stantiating i from 1 to n in the description.

Deﬁnition 3. For a given AnalysisProject p =

(t, wd, rd, gl), one deﬁnes the following objects.

• Enriched Text: Crossing the text t and wd, the en-

riched text proposed a version of the text in which

the words that appear in wd (or whose lemma ap-

pears) are tagged and associated with the same

colour as in wd. Formally,

Enriched Text :

Text × P(FieldWord) → TaggedText

Enriched Text(t, wd)

= Append

lengt(t)

i=1

coloured(t

)

where

coloured(t

)

= t

< wd

.colour >

if t

∈ deriv(wd

.word)for any j,

otherwise

• Field Dictionary: enriching the world dictionary

wd with knowledge extracted from the text t, the

ﬁeld term dictionary is the list of purposeful words

associated with their colour and a natural number

denoting the number of occurrences of the word

in the text (or the number of its declination if the

word is a lemma). This is formalised as follows.

Field dictionary :

Text × P(FieldWord)

→ P(FieldWord ×Colours × R)

A Formal Model to Support Discourse Semantic Landscape Analysis

121

Field dictionary(t, wd)

= Append

length(wd)

i=1

.word, w

.colour,

#{w ∈ t : w ∈ deriv(wd

.word)})

• Field Relation Dictionary: the relation dictionary,

rd, where each word is augmented with the colour

of the word in wd and a Boolean denoting the

presence/absence of the word (or one of its deriva-

tive) in the text. Formally

Field Relation Dictionary :

Text × P(FieldRelation)

→ P(FieldWord × Boolean × FieldWord

×Boolean)

Field dictionary(t, rd)

= Append

length(rd)

i=1

.w1, (∃w ∈ t : w ∈ deriv(r

.w1)),

.w2, (∃w ∈ t : w ∈ deriv(r

.w2)))

• Field Relation Matrix: the symmetric matrix M of

dimensions |wd|×|wd| where each element m

i, j

a colour that indicates the existence and the type

a relation (wd

, wd

) ∈ rd. The possibles values

are red for an opposition, green for an association

and grey if no relation. Formally, denoting Col the

set {Red, Green, Grey},

Field Relation Matrix :

P(FieldWord) × P(FieldRelation) → M (Col)

Field dictionary(wd, rd)

= {m

i, j

= Red iff (wd

, wd

, 0) ∈ rd,

Green iff (wd

, wd

, 1) ∈ rd,

Grey otherwise}

• Visualisation which is a graphical representation

of a DomMap ∈ gl enriched with information in-

volved in p as the colours associated with words,

their occurrence in the text and the relations be-

tween the words.

Direct access to these enriched objects provides to

the human scientist an enriched perception of the text,

and guides him into the interpretation process.

At this stage, for the sake of genericity, the model al-

lows any kind of domain for visualisation. In the follow-

ing we describe more precisely some of the mapping imple-

mented in Evoq.

4.3 Project Transformations and

Analysis Project Construction

The model presented above formalises a post-

structuralist enhanced model of a text, as the result

of an analysis, and the enriched views on the infor-

mation. We describe in section 5 how it efﬁciently

serves the intermedial transposition. At the formalisa-

tion level, a question remains to address: how to build

such a model from a fresh text? Expressed in our

formalism: how to transform an original (t,

0, k)

into an analysis (t, wd, rd, ml, k) and its associated en-

riched objects?

Let us remind that we do not aim to offer a tool

that will fully automate the analysis but a tool that

will, on the one hand, stimulate the analyst and, on

the other hand, facilitate some tasks of encoding or

research. That is to say that only a part of the knowl-

edge k requested to lead the analysis can be fully for-

malised and implemented, an important part of it re-

mains in the analyst’s brain.

With this in mind, the analysis process can

be formalised as a sequence of transformation that

will, step-by-step, transform a analysis project ﬁeld

(t, wd, rd, ml, k), and

• extract words from the text and/or the knowledge

using the text and/or the (integrated or human)

knowledge, and

• add these words, with their associated colour, to

wd,

• identify relations among the words, from the text

and/or the knowledge, and

• add these relations, with their associate boolean,

to rd.

• create or adapt spatial manifestations of the

words and the relations to express semantic ﬁelds

and their tension.

Whether the operations are performed by a human an-

alyst or proposed by an AI agent, our aim here is to

show that they can be supported by the proposed for-

mal model.

The current version of the Evoq platform offers

a basic function allowing the extraction of the most

frequent words from the text, with or without the ex-

clusion of stopwords. The version including all words

can be modelled by

FrequentWords(t, wd, rd, ml, k) : AP → P(FieldWord)

t → {(w, black) : #{i : t

∈ deriv(t

)} ≥ m}

where t

denotes the ith word of the text t, m denotes

the arbitrary minimal number of occurences k in-

volves the capability to retrieve derivatives of a word.

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

122

And the one removing stopwords by

FrequentKeyWords(t, k) : Text → P(FieldWord)

t → {(w, black) : #{i : t

∈ deriv(t

)\Stopwords} ≥ m}

with the same conventions, and Stopwords being a

list of stop words involved in k.

The addition of a (colored) word (w,col) is de-

ﬁned only if w is not already a member of wd

addWord : ((t, wd, rd, ml, k), (w, col))

* (t, wd ∪ {(w, col)}, rd, ml, k)

The addition of relation is the direct adaptaion of

the one on the basic model and similarly deﬁned only

if one of (w

, w

, 0) or (w

, w

, 1) is in rl:

addRel : ((t, wd, rd, ml, k), (w

, w

, b))

→ (t, wd ∪ {(w

, w

}, rd{(w

, w

, b), (w

, w

, b), ml, k)}

The operations for correction in an analysis,

namely, remove a term or a relation follow also di-

rectly from these on the basic model.

removeWord : ((t, wd, rd, ml, k), w)

→ (t, wd \ {(w, col) : col ∈ Colours},

rl \ {(w

, w

, b) : w

= w ∨ w

= w}, ml, k)

removeRel : ((t, wd, rd, ml, k), (w

, w

, b))

→ (t, wd, rd \ {(w

, w

, b), (w

, w

, b)}, ml, k)

A last operation consists in modifying the colour

of a word already present in the word dictionnary, this

is formalised by the partial function

changeColour : ((t, wd, rd, ml, k), (w,col))

* (t, (wd \ {(w, c) : c ∈ Colours}) ∪ {(w, col)},

rd, ml, k)

5 SEMANTIC LANDSCAPE:

INTERMEDIAL

TRANSPOSITION

In this section we highlight how our formalisation of-

fers a unifying model and supports the intermediary

transposition between different visualisations associ-

ated with an Analysis Project which constitute its se-

mantic landscape.

This section is illustrated with the analysis of

the text available on (Transition Network, 2021) that

presents the principles and value guiding the move-

ment called Transition Network. As mentioned above

the basic element of this analysis are the text t, the set

of 21 words wl, and a list of 52 relations rl.

In the intermedial model, words are associated

with colours in the word dictionary wd, the relation

dictionary rd = rl. The list of mapping ml and ele-

ments of the knowledge k that complete the Analysis

Project AP = (t, wd, rd, ml, k) are described below.

After a literature review of text-related visualisa-

tion, (Clarinval et al., 2018) proposed a selection of

visuals and studied their respective suitability to sup-

port the structural analysis.

This section present in turn, enriched text, words

and relation dictionaries, chord diagram, adjacency

matrix and node-link diagram.

5.1 Enriched Text

The text presentation built into Evoq provides a di-

rect transposition of the Enriched text deﬁned by the

Enriched Text function in the subsection 4.2. To

provide this visualisation, the domain mapping con-

structs the enriched text and then calls a function that

transforms each tagged word into the html expression

that applies the required formatting.

Figure 1 illustrates the visualisation of the begin-

ning of the text of our running example

Figure 1: Enriched text visualisation.

5.2 Dictionaries

Dictionaries can be seen as simple lists and presented

as such. However, as they are not isolated objects but

elements of a complete analysis project AP, the Evoq

software propose enriched view of the dictionary.

Namely, the word dictionary wd is completed with

the number of occurrences of each of its words in the

text t using the Field dictionary function and the re-

lation dictionary rd is completed with information re-

lated to the words using the Fied relation dictionary

function.

Here again, the mappings consist in computing the

enriched elements and then turning the results into

The text is an extract from (Transition Network, 2021)

used to build our running example

A Formal Model to Support Discourse Semantic Landscape Analysis

123

Figure 2: First items of Word Dictionary Visualisation.

Figure 3: First Items of Relation Dictionary Visualisation.

HTML code that construct the tables. Figures 2 and 3

respectively present the ﬁrst items of the word dictio-

nary and relation dictionary of our running example.

Note that, in ﬁgure 2 the absence of a word in the text

is revealed both by the null number of occurrences

and the parenthesis surrounding the word. Similarly,

in ﬁgure 3, the words are presented with the colour

deﬁned in wd and are also surrounded by parenthesis

if they are absent from the text t.

5.3 Adjacency Matrix

Relations being at the heart of the structural analy-

sis, several visuals are proposed from the relation dic-

tionary. The adjacency matrix proposes an enriched

vision from the Field Relation Matrix function de-

ﬁned in subsection 4.2. After computing the Field

Dictionnary and Field Relation Matrix, the mapping

create an html tabular having the coloured words as

columns and lines headers and each cells receiving

the coloured (red/green/grey) deﬁned by the Field Re-

lation Matrix corresponding cell, but for the diagonal

elements which are dark grey coloured.

Figure 4: Top-left corner of the adjacency matrix.

5.4 Node Link Diagram

A last visual associated with the set of relationships

is to present the relation dictionary as a node-link di-

agram. For this visual the mapping consists of

• projecting each word in the dictionary into a two-

dimensional space

• writing the words in the corresponding colour and

surrounding them with brackets if necessary, ac-

cording to the convention explained above

• draw the arcs corresponding to the relations, in red

for oppositions, in green for associations.

After the computation of the Field Relation Dictio-

nary, the last two aspects do not pose any critical

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

124

problems. The question of the placement of points

in space is much more delicate. A large literature

deals with the question of the placement of the points

of a graph in particular with the objective of limit-

ing the crossings of the arcs. The dictionaries of re-

lations integrating two types of relations, these so-

lutions proved to be unsatisfactory in terms of inter-

pretability of the produced diagram. The Shock Wave

algorithm (Cauz et al., 2021) has been speciﬁcally de-

veloped to visualise relation dictionaries. It places the

nodes to highlight the structural axes represented by

the opposition relations.

Figure 5 presents the node-link diagram created

for the Transition example.

Figure 5: Node-Link diagram.

Each of the visualisations integrates its own inter-

action mechanisms which give access to the functions

described in the section 4.3 : add, delete or modify

a word, add or delete a relation (for more details see

(Clarinval et al., 2018)). The node-link diagram also

offers the speciﬁc possibility of modifying the map-

ping function by changing the position of a node with

a simple drag and drop.

6 CONCLUSION

The qualitative analysis of unstructured texts is an es-

sential task for human science researchers. It remains

largely unexplored by the IT world, which offers these

researchers essentially support tools for the manipula-

tion of texts and elements of analysis that do little to

enrich the analyst’s reﬂection or focus on quantitative

approaches (Lejeune, 2010).

Through the contribution presented in this work,

we develop an innovative perspective which aims to

propose a formal modelling of the analyst’s approach

in order to offer him a variety of visual supports that

multiply the views on the analysis in progress while

preserving their coherence.

To achieve this ambition, the formal modelling of

the analysis integrates in its structure the theory of

language mobilised for the analysis: structural analy-

sis. This paper proposes a formal model of knowledge

that incorporates this theory. It also illustrates how

this model allows for intermediary transposition and

ensures the overall consistency of the semantic land-

scape. The transpositions that make up the semantic

landscape provide rich tools both for communicating

analyses and for stimulating the creativity of the re-

searcher.

This work served as the basis for the development

of the Evoq tool (Clarinval et al., 2018; Linden et al.,

2020) available on https://evoq.info.unamur.be/login.

REFERENCES

Cauz, M., Albert, J., Wallemacq, A., Linden, I., and Dumas,

B. (2021). Shock wave: a graph layout algorithm for

text analyzing. In Patrick Healy, M. B. and Bonnic,

A., editors, Proceedings of the 21st ACM Symposium

on Document Engineering. ACM Press.

Clarinval, A., Linden, I., Wallemacq, A., and Dumas, B.

(2018). Evoq: a visualization tool to support structural

analysis of text documents. In Proceedings of the 2018

ACM Symposium on Document Engineering, United

States. ACM Press.

Derrida, J. (1967). L’

ecriture et la diff

erence. Editions du

Seuil, Paris.

Ellestr

om, L., editor (2010). Media Borders, Multimodality

and Intermediality. Palgrave Macmillan, London.

Ellestr

om, L. (2014). Media transformation. The Trans-

fer of Media Characteristics among Media. Palgrave

Macmillan, London.

Gibson, H., Faith, J., and Vickers, P. (2013). A survey

of two-dimensional graph layout techniques for infor-

mation visualisation. Information Visualization, 12(3-

4):324–357.

Kucher, K. and Kerren, A. (2015). Text visualization tech-

niques: Taxonomy, visual survey, and community in-

sights. In Proceedings of IEEE Paciﬁc Visualization

Symposium (PaciﬁcVis), pages 117–121. IEEE.

Lejeune, C. (2010). Montrer, calculer, explorer, analyser. ce

que l’informatique fait (faire)

a l’analyse qualitative.

Recherches Qualitatives, 9:15–32.

Lejeune, C. (2021). http://www.squash.ulg.ac.be/lejeune/.

web page dedicated to Qualitative Data Analysis Soft-

wares, accessed:2022-03-29.

Linden, I., Wallemacq, A., Dumas, B., Deville, G., Clarin-

val, A., and Cauz, M. (2020). Text as semantic ﬁelds:

Integration of an enriched language conception in the

text analysis tool evoq

. In Fabiano Dalpiaz, J. Z.

and Loucopoulos, P., editors, Research Challenges

in Information Science - 14th International Confer-

ence, RCIS 2020, Proceedings, volume 385 of Lec-

A Formal Model to Support Discourse Semantic Landscape Analysis

125

ture Notes in Business Information Processing, pages

543–548.

Rajewsky, I. (2002). Intermedialit

at. A.Franke UTB,

Stuttgart.

Rajewsky, I. (2005). Intermediality, intertextuality, and re-

mediation: A literary perspective on intermediality.

Interm

edialit

es / Intermediality, (6):43–64.

Transition Network (2021). https://transitionnetwork.

org/about-the-movement/what-is-transition/

principles-2/. accessed:2022-03-30.

Wolf, W. (1999). The Musicalization of Fiction. A Study in

the Theory and History of Inter mediality. Amsterdam

; Atlanta, GA : Rodopi.

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

126