Abstract
There is an increasing use of information visualization techniques in Digital Humanities to support the analysis
of literary works. On the other hand, the rise of digital reading on the web as well as the development of dedicated
e-book readers has triggered a variety of hardware and software solutions to provide new reading environments and new
material engagements with textual forms. In this paper, we propose a new metaphor, In-Out-In metaphor, to support the
reading of literary works in a digital medium. Our model integrates information visualization techniques with plain text
reading into the reading flow. This approach differs from the well-known dichotomy between close and distant reading
because its emphasis is not on supporting the analysis of the texts being read, but on providing a smooth flow between
focus and digression, which naturally occurs during a reading experience. We present a solution to read and explore the
Book of Disquiet by Fernando Pessoa, an unfinished book composed of a set of modular texts that have been edited in different
sequences and which can be read multi-sequentially. The research and design decisions for the visual reader of the Book of Disquiet
are meant to provide an appealing, rich and interactive online reading experience through a web-based application, LdoD Visual.
The implemented features apply information visualization techniques with an emphasis on exploring the modular nature of the book.
Since this is a fragmentary literary work, its reading can be fragmentary as well, but providing a smooth multiple flow reading experience.
1. Introduction
The
Book of Disquiet (LdoD, from
Livro do
Desassossego, in Portuguese) by Fernando Pessoa is a fragmentary literary
work that was left unedited and unpublished by the author. In 1982, forty seven
years after Pessoa’s death (1935), its first edition was published
[1]. Since then, three additional critical editions
of the book have appeared, showcasing significant disparities of interpretation
about how the book should be organized. Between 2012 and 2017, a team led by two
co-authors of this article (Manuel Portela and António Rito Silva) created the
LdoD Archive: A Collaborative Digital Archive of the Book of
Disquiet [
Portela and Rito Silva 2017]. The
LdoD
Archive can be described as a hypermedia interactive engine that has
realized the vision of an endlessly reconfigurable book. It is both a scholarly
digital edition and meta-edition that enables users to compare each version of the
Book of Disquiet with all of the versions encoded in the
archive, and a virtual editing space designed with game-like principles for enabling
different types of users (experts, students, general readers) to make their own
editions and collections of the
Book of Disquiet. The stated
aim of this evolving archive is to experiment with computational forms of editing,
reading, and writing in ways that go beyond the bibliographic imagination
[
Portela and Rito Silva 2015][
Rito Silva and Portela 2015][
Portela 2017][
Portela 2019][
Portela 2022]. The
LdoD Visual – a web application integrated into the
LdoD Archive – explores the reading flow within this complex
digital environment.
The modularity of The Book of Disquiet is, in fact, the
contingent result of material and historical features that tend to reinforce each
other: on one hand, each text is a self-contained unit that has no necessary
narrative continuity with other texts; on the other hand, there is no defined
ordering of the entire corpus of circa 600 pieces. Autograph textual units are also
in various stages of revision: first draft manuscripts, typescripts with manuscript
emendations, clean typescripts, twelve published texts. Pessoa only defined a few
textual clusters, and even those suffered changes over time during the two stages of
composition (1913-1920; 1928-1934). Given this contingent modular nature of the
book, its fragments can be read regardless of any particular order, offering the
opportunity for a fragmentary reading of the book, that is, a practice of reading
that fosters divergent reading paths that can be built by the actual reader, instead
of following a typical sequential reading that is bound to happen in a rigidly
defined story sequence.
In the
LdoD Visual, we used the particular structure and
characteristics of the
Book of Disquiet to explore, and
smoothly integrate, two current research trends in Digital Humanities: information
visualization techniques applied to texts, including the possibility of moving
between close and distant reading scales [
Jänicke et al. 2015], with the design of
web interfaces for accessing complex literary archives and the emergence of e-books
as a reading medium [
Koolen et al. 2012]. We propose a new reading metaphor, the
In-Out-In metaphor, which integrates information visualization techniques with the
plain reading of the text, to support the focus-digress flow that structures the
embodied activity of the reader. By focus-digress flow we
mean that during the focus stage the reader is in the story, while the digress stage
corresponds to moments of self-consciousness of textual form, structure and
navigation, as well as associative thought processes, both of which interrupt
immersive reading.
We have designed and implemented a web-based application to read the
LdoD whose main objective is to give the reader a sense of choice and an
active role while reading and browsing through the book’s multiple textual
fragments, in what should be a pleasurable and immersive user experience for both
humanities researchers or just interested readers. The entire rationale of the
LdoD Archive has been to offer a multi-layered access to the
text, mediated through specific interfaces for each major function. Interfaces
should be meaningfully navigated by interactors with very different levels of
knowledge of the text – from the novice to the learner to the expert – and with
distinct aims in mind, including leisure reading, study and analysis, advanced
research, and literary creation [
Portela 2017].
To achieve this goal at the level of the reading interface, we integrate a set of
meta-information about the
Book of Disquiet with information
visualization techniques. Information visualization is a broad field that can be
applied to a large domain of different areas, providing insight based on input data
with different perspectives, summarizing relationships, identifying and comparing
patterns, trends, outliers, finding correlations, among other tasks [
Manovich 2011].
In this project, visualization has been conceived as a humanistic interface for
integrating different reading strategies and proving readers with a perception of
reading as the traversal of a semiotic space that is both topological and semantic,
and which is responsive to the readers’ agency [
Drucker 2014][
Murray 2012].
In order to integrate the existing meta-information with information visualization
techniques we developed a tool, the LdoD Visual, which
supports the exploration and pleasurable reading of the Book of
Disquiet. The implemented solution provides new contributions for the
reading and exploration of this work, in particular, and to the concept of reading
flow in Digital Humanities, in general. This tool is built on top of the Collaborative Digital Archive of the Book of Disquiet, as
mentioned. So the visual reader should be understood as a second-degree interactive
representation – i.e., an additional interface – for providing a rewarding reading
experience of a highly complex corpus of 600 texts. Its conception brings together a
metaphoric model of the reading process, on one hand, and an algorithmic and
graphical modelling of that process, on the other.
Besides describing the technical and graphical details of the
LdoD
Visual as a reading interface, this article makes three related claims. We
argue that our computational model of reading interactions is adequate for
representing the multiplicity of reading traversals inherent in the
Book of Disquiet as a modular and open network of textual
sequences. We also claim that our model of the reading flow as a recurrent process
of immersion-emersion can be equated with the networked digital medium itself,
insofar as the continuous focus on strings of text required for immersion can be
smoothly integrated with the macrotextual logic of the visualizations facilitating
textual overviews, hypertextual connections, and frame jumps. Finally, we argue that
the ways for moving into, moving within, and moving out of the text provide a
satisfactory balance involving two scales of textual representation that could be
useful for other digital humanities projects of highly complex textual corpora. We
believe the ergonomic solution embodied in our reading interface explicitly
integrates close reading, hyper-reading, and machine reading [
Hayles 2012], and thus
contributes to the design of new reading interfaces.
2. Related Work
In this section we present some background concepts regarding the LdoD Archive, Visualization in Digital Humanities, and Classification in
Digital Humanities, the three conceptual and technical contexts in which this work was developed.
2.1. The LdoD Archive
The
LdoD Archive provides every reader with the possibility of
reading the
Book of Disquiet according to
four expert editions: the first critical edition, edited by Jacinto do Prado
Coelho [
Coelho 1982], which was published in 1982, as well as three critical versions that have been edited by Teresa Sobral Cunha (1990-91 [2008] [
Sobral 2008]),
Richard Zenith
(1998 [2012] [
Zenith 2012]) and Jerónimo Pizarro (2010) [
Pizzaro 2010]
[2][3].
Each edition is composed of the total number of fragments and their relative order varies from edition to edition. In the LdoD Archive, the user can read each fragment while comparing
its respective position across these expert editions, and they can also have access
to additional information, which may also vary according to the edition, such as
date and heteronym attribution.
In the LdoD Archive, users can also create their own virtual
editions of the LdoD by selecting fragments from each of the 4
expert editions, sorting, annotating and tagging them. This virtual edition creation
can be produced in a collaborative manner, in which case several users are involved,
and users can play the roles of editors and managers of a virtual edition. The
tagging of a fragment, which can apply to the fragment as a whole, or to just a
segment of the fragment, is done by means of a set of categories that is associated
with the virtual edition. This set of categories is designated as the virtual
edition’s taxonomy.
Each fragment has different properties, such as its title, text, date and the edition
and categories to which it belongs. There are many more features and possibilities
that can be explored in the archive, such as additional reading tools, original
documents, comparisons between textual transcriptions, fragment search, multiplayer
classification games, and the writing of variations based on the existing
fragments.
2.2. Literary Reading on the Web
Whereas hypertext structures in print books are generally subordinated to the
sequential flow of textual strings on the page (they gravitate around the text, as
table of contents, footnotes, references, etc.), hypertext structures in digital
media alter the dynamics between self-contained and associative reading, as they can
place any textual string in direct relation to a multiplicity of other texts, thus
increasing the ratio of textual interruption. Links among disparate strings of text
become much more frequent, and readers follow networks of association in interrupted
and labyrinthine traversals. With the emergence of the World Wide Web as a
hypermedia environment in the 1990s and early 2000s, the integration of verbal text
into multimodal genres of communication continued to transform our media for
symbolic production and exchange. More recently, mobile social media and the
development of smartphones as multifunctional devices further extended this general
transformation of digital literacy.
Even if visionary engineers imagined the future interface of computational devices as
a sort of augmented book (as is clear from Alan Kay’s 1972 Dynabook prototype), and
even if current electronic tablets somehow attempt to emulate the ergonomics of the
portable book, the electronic writing and reading space reconfigures and reforms the
set of relations inherited from bibliographic structures. The redesign of
information spaces of legacy print forms for the networked screens of programmable
media has been a source of continuous reflection and engineering experimentation in
many areas – from scholarly editing to commercial publishing to educational tools,
particularly since the late 1990s. These transformations have also been studied
extensively, from human-computer interaction to studies of literacy to cognitive
science [
Baron 2015][
Mangen et al. 2016][
Mangen et al. 2019]. Studies seem
to concur on the significance of changes to the ecology of reading practices, and on
cognitive differences in terms of orientation and comprehension. Complex hypermedia
literary archives, such as
The William Blake Archive
[4] (under development for
almost three decades), have had to adapt to this changing ecology of the web. In
those scholarly editing projects combination of search functions, navigation
structure and reading interface has undergone several redesigns from highly
hierarchical networks of links – typical of the early hypertext remediation of
bibliographic structures – to current on-the-fly aggregations based on specific
XML-encoding conventions.
Interviewed by Werner Herzog, in
Lo and Behold:
Reveries of the Connected World (2016), Ted Nelson recalls a childhood
episode when he was with his grandparents in a row boat, and, trailing his hand in
the water, he notices the water flowing around his fingers in intricate patterns of
connection. This epiphanic moment gives him awareness of the connectedness of
everything which he associates to his future vision of hypertext as a powerful
open-ended multidirectional changing system of relations among all texts and all
media. Non-sequential hyperlinked literature as an
evolving network of transclusions challenges reading practices based on linear
forms of writing [
Nelson 1965][
Nelson 1995][
Nelson 2003].
Although not as flexible and unconstrained as imagined by Nelson, the current
mediascape of reading was radically transformed by hypermedia and algorithmic
textual production and analysis. Despite the early apology for the liberating,
dynamic and heterarchical potential of hypertext by literary theorists and digital
writers [
Moulthrop 1991][
Bolter 2001][
Landow 2006], research into hypertext software
had always recognized the disorienting effect of link and node structures as both a
design and cognitive problem [
Theng et al. 1996][
Theng and Thimbleby 1998], to which alternative
representational and navigational forms, such as spatial hypertext, have been
proposed [
Marshall et al. 1995][
Bernstein 2011][
Solís and Ali 2011].
Almost three decades after the first literary experiments with hypermedia, Stuart
Moulthrop and Dene Grigar (
2017) have asked four hypertext writers (Judy Malloy, John
McDaid, Shelley Jackson, and William Bly) to revisit their early work of the late
1980s and 1990s in order to recover the writing, programming, and reading mechanics
involved in their exploratory uses of specific machines and programs, including the
emerging conventions of the hyperlink as a resource for poetic and narrative
expression. Moulthrop and Grigar’s deep reading of the works of those authors
engages the multilayered nature of digital textuality, including the ways in which
hypertext and algorithmic processes affect our reading expectations of texts as
holistic self-limited entities. The elusiveness of the textual whole is inherent in
the intricate topologies created not just by hyper-reading as an unbounded and
labyrinthine traversal, but also by the factorial virtualization of text as a
responsive permutation of language bits.
Many of the same cognitive challenges of early digital genres continue
to be addressed in the design of dedicated e-book readers and in current attempts to
create a satisfactory balance between competing reading modes in digital media.
Several analyses have been done on the hardware and software features for reading in
electronic environments [
Koolen et al. 2012] that highlight the differences between
linear sequential reading and multicursal interactive reading, which are
particularly relevant in the context of academic research or education. This tension
between continuity and discontinuity pervades all solutions, in which there is a
tradeoff between the smooth flow of reading and the possibility of analysis and
inquiry about what is being read. If we think of Digital Humanities as the coming
together of digitized artifacts, digitized methods for modelling and analyzing those
artifacts, and digitized interfaces for representing those artifacts and those
analyses, then the development of reading interfaces remains an important component
in the invention of rhetorical conventions and modes of communication in networked
programmable media.
2.2.1. Reading in the LdoD Archive
Prior to this work, the
LdoD Archive provided two reading user
interfaces: one that emphasizes
comparison
between editions and another that emphasizes the sequence of
reading
(Figures 1 and 2).
In the comparison user interface, it is possible to have a detailed highlighting of
the differences between the authorial sources and the editorial transcriptions, and
also across the editorial transcriptions themselves. It is also possible to annotate
particular texts (or textual passages) with comments and tags when specific texts are used in the creation of a virtual edition.
The differences between transcriptions among the different expert editions, and
authorial sources, are highlighted by placing the fragments side by side, and
using colors to identify which parts differ. This user interface also allows the
comparison of meta-information assigned by the expert editors, such as title, date
of publication, heteronym attribution, page number, which can differ according to
the edition. Additionally, this interface also allows users to compare the
transcription with the original document sources, by presenting the documents’
facsimiles and providing a diplomatic transcription, which includes deletions,
insertions and substitutions.
On the other hand, and since the LdoD Archive also supports
the definition of virtual editions, it is possible, through the comparison user
interface, to compare the differences between the fragments, in terms of the
assigned tags and annotations done. It is through the comparison interface that
authenticated users can annotate and tag the fragments.
Due to the focus of the comparison user interface on the analysis aspects, another
user interface was defined, in order to provide a smother reading experience,
although still emphasizing the comparison between the different sequences of reading
the LdoD. This interface has a look and feel closer to the
printed typography of the period the LdoD was written, and
intends to provide reading sequences according to each one of the four expert
editions. Besides, it provides a recommendation interface that suggests the next
fragment to read, given the fragment being read and a set of criteria, such as, for
instance, text similarity between fragments.
We could say that the reading interface for comparing editions is aimed at scholarly
engagements with the text, whereas the reading interface for showing divergent
reading sequences is meant for general readers. In the latter interface readers have
the option of reading vertically within a particular sequence or horizontally across
the entire corpus, thus redefining their textual traversal according to the
multicursal logic of the medium which, in turn, is critically used to explore the
modular structure of the work itself. On the other hand, the aim of the new visual
reader is to abstract textual units and their meta-information and present them
visually in ways that enhance movements within and across the text. The hypertextual
and textual dynamics of the LdoD Visual embodies an ergonomic
and cognitive model of the reading process in the design of the interactions.
2.3. Visualization in Digital Humanities
Jänicke et al. (
2015,
2017) address visual text analysis in Digital Humanities and
categorize each of these techniques as distant reading techniques, which aim to
generate an abstract view of textual content, and as close reading techniques, which
rely mainly on annotations and the use of different colors and underlining styles to
lead to a deep comprehension and thorough interpretation of textual passages. They
analyze how close and distant reading techniques can be combined in ways that could
bridge these two perspectives so that the user can switch between close and distant
reading. This is mainly achieved by side-by-side visualizations of the text being
read and its meta-information, which make it possible to switch from a close reading
of a particular fragment to some kind of overview associated with specific metadata.
On the other hand, it is possible to access a particular text from the distant
reading view where the text is represented because it has properties that are
analyzed in that distant reading mode [
Cheema et al. 2016]. This integration between
close and distant reading is similar to our goal but our emphasis is on providing a
reading experience, while the integration of close and distant reading is concerned
with the analysis and understanding of the texts, such that close reading is
frequently related with the annotation and generation of meta-information while the
user is reading a text.
2.3.1. Visualization Techniques for LdoD
Given the specific characteristics of the LdoD we can analyze
which visualization technique can be applied in its representation. These techniques
belong to different areas and types of tasks, such as reading guidance, user
navigation, global view, and exploratory analysis.
Reading guidance features consist of giving important information for the user
regarding a specific fragment or even a collection or category of fragments, where
the focus is on the fragment that is currently being read. User navigation features
provide the user with an interface to navigate between different parts of the LdoD, in which the focus is on the relationship between the
fragment being currently read and the remaining fragments of the book. Global view
features have the objective of giving the user a summarized presentation or
zoomed-out view of content from the LdoD, thus directing the
focus to the edition that is being read. Exploratory analysis features present the
user with information that is more focused on exploring relationships in the LdoD without necessarily converging to the task of
reading.
Word clouds [
Viégas et al. 2009][
Mohammad 2012][
Heimerl et al. 2014] can be used to
provide global views and reading guidance. Word clouds can be associated with the
taxonomy categories of a virtual edition. After being presented with a word cloud of
the available categories, and clicking on one of them, the user would be redirected
to a list of fragments that corresponds to the selected category.
Also by exploring the categories associated with the virtual editions, it is possible
to provide a global view of the edition, using colors to identify those fragments
associated with a particular category. This feature would provide a global view of
the fragments tagged with that category and the navigation to another fragment that
is associated to the same category. This visualization technique, is, for instance,
used in the Diggersdiaries web site [
Vilaplana and Pérez-Montoro 2017].
Besides the categories criterion, other criteria are supported by the
LdoD Archive, such as text and heteronym, and they can be
used to define a distance value between fragments. In this case, when reading a
given fragment, the user could browse through similar fragments according to the
selected criteria, and this relation can be visually represented using a network
graph to show how close certain fragments are in relation to others. This would
provide reading guidance. Alexander and Gleicher (
2016) make use of network graphs
to visualize text similarity between different work by William Shakespeare.
By highlighting some of the fragments’ words according to their relevance in the
context of the entire book it is possible to provide a guidance while reading a
fragment. The weight of this highlighting could be made resorting to a metric like
Term Frequency – Inverse Document Frequency (TF-IDF)
[5].
Text Skimming [
Brath and Ebad 2014][
Brath and Ebad 2015] is a
technique used in knowledge maps and to support reading and it can also be applied in
this context.
There are some differences between the experience of reading a book on a browser or
other portable virtual book readers and the actual physical book. In an attempt to
close this gap, it would be interesting to provide the user with a graphical
representation that makes the awareness of where the actual reading fragment is
placed in the current edition. This type of feature belongs to the global view
area.
2.4. Text Classification
The visualization techniques are built on top of the information to be presented plus
some meta-information. In the context of the LdoD Archive,
the information is the fragment transcription and its position in an edition, and
the meta-information is the date, heteronym and categories assigned to a fragment.
Additionally, some meta-information is generated from the fragment through the
TF-IDF associated with it, by considering each fragment as a document and the words
in the fragment its set of terms.
2.4.1.Distance Measures in the LdoD Archive
The
LdoD Archive implements a set of measures to calculate the
distance between two fragments belonging to the same edition according to four
types of meta-information: date, heteronym, categories, and TF-IDF. The distance
between two fragments is calculated using the cosine similarity [
Singhal 2001],
according to which a vector is defined for each of the fragments to be compared.
[6]
Associated with the heteronym, a vector of two cells is defined, one for each
heteronym, Vicente Guedes and Bernardo Soares, in which the value 1.0 is assigned to
the cell of the heteronym if the fragment is assigned by the editors to it.
For the date, a vector is defined with the number of cells equal to 1934-1913+1, which
corresponds to the period when the fragments were written. For the cell
corresponding to the year when a fragment was written, the value 1.0 is assigned and
for the other cells a decay of MAX(0.0, 1.0 - N x 0.1) is applied where N is the
distance between the year associated with the cell and the closest cell with value
1.0.
For the text, the set of the first 100 TF-IDF terms associated with each one of the
fragments that we intend to compare is calculated. Note that this set is between
100, in which the two fragments have the same first TF-IDF 100 terms, and 200, in
which all terms are different. Then, for each fragment, a vector whose size equals
the number of terms in the set is defined, and to the cell associated with a term
the value 1.0 is assigned if the term belongs to the first 100 TF-IDF terms of the
fragment, and 0.0 otherwise.
To calculate the distance between two fragments using the set of categories assigned
to them a vector whose size is equal to the number of categories associated with the
virtual edition taxonomy is defined. Then, for each fragment, its respective vector
is filled with 1.0 for the cells corresponding to the categories the fragment has,
and 0.0 for the other cells.
Note that the archive also uses a topic modelling algorithm, through the Mallet
software
[7], which
allows the virtual editions to be automatically classified using the set of
generated topics.
3. The In-Out-In Metaphor
Reading a book is a task in which the reader constantly switches between two states
of focus. The reader is either completely engaged in the reading task or briefly
disengages out of the act for various reasons. These moments of disengagement can be
triggered, for example, by recalling what happened in previous chapters, picturing a
description of a character or scenario, remembering a similar personal experience in
relation to what has been read, linking a new event to what has happened in the past
or even trying to make a prediction out of it, among other possibilities. Immersion
can also be interrupted by the readers’ self- awareness of form, structure and
navigation. The in-moment represents the state of being inside the textual world,
absorbed by its signifying chain. The out-moment points to the liminal state of
being outside the text, a liminality that is a function of the text itself and the
ways in which it draws attention to its materiality or generates mental processes of
association and reflection.
Part of our solution model revolves around this phenomenon, as it is intended to
materialize and direct this constant in and out that happens while emerging out of
and submerging into the source text. It is important to highlight that, in
comparison to a typical book where we have a rigidly defined story sequence, the LdoD is a fragmentary literary work. This means that these
moments of disengagement can turn into a chance for the user to reposition herself
in another part of the book. This nuance leads to a reading path and global mental
picture of the world created by the book that can greatly vary from reader to
reader, opening up space for providing the user with an immersive experience in
which she can have a sense of choice and an active role while reading the LdoD. If we consider that genre is an emergent property that
results from specific arrangements of textual units, readers of the Book of Disquiet - mediated by this in-out-in visual reader –
may even experience the oscillation between fictional autobiography and fragmentary
narrative as an effect of their traversals. The emergence of a variable textual
whole becomes a function of their reading acts. Self-contained textual units, when
traversed as reading text, correspond to the in-movement, but when navigated as
visual representations correspond to the out-movement.
4. The LdoD Visual
The solution’s interface is composed of a group of different main menus: edition
selection, reading, current user activity, new user activity, and reading history
menus. Before describing each menu in detail, it is important to explain the
concepts that will be used throughout this solution proposal.
4.1. Core Concepts
An information visualization technique, among other
descriptions in previous sections, is a visual representation or metaphor that can
provide insight and different perspectives about almost any type of input data, as
we saw above from the multiple examples given in the related work applied to the
context of the Book of Disquiet.
Semantic criteria is the set of attributes that can be encoded
in the information visualization techniques in the context of the Book of Disquiet, namely word frequency, word relevance (TF-IDF),
chronological order, heteronyms, fragment categories, and taxonomies. Text
similarity is another semantic criterion that can be derived from word frequency and
word relevance.
A user activity is an activity that can be selected by the
user while reading a fragment from the Book of Disquiet,
consisting of making use of an information visualization
technique to encode different semantic criteria. An
example of a user activity is “Find similar fragments by text similarity” to find
fragments that are similar to the fragment that is currently being accessed, using a
network graph as the information visualization technique that encodes text
similarity - the semantic criterion that is chosen by the user while selecting the
new user activity.
To achieve an intuitive and easy-to-use solution, independently of the user
activity's information visualization technique and encoded semantic criteria, the
user selects a user activity by clicking on a button with a high-level description
of what the user activity was designed to achieve, e.g. “Find similar fragments by
text similarity”.
A reading flow is a sequence of accesses, done by a single
user, such that the user progresses in the sequence of fragments through in-in, in-out-in and in-out-out-in subflows. An in-in subflow
corresponds to a situation when the user is accessing a fragment and accesses
another fragment, for instance, by switching to the previous page. This switch is
done according to the semantic criteria the user chose for
accessing the fragment, defined by the current user activity. For instance, if the
semantic criteria correspond to the set of fragments of a given category, the reader
will access another fragment belonging to the same category. An
in-out-in subflow corresponds to a situation in which the user is
accessing a fragment, moves to the current user activity menu (to be described in
the next subsections) in which the reader has a complete view of the fragments
according to the same semantic criteria used to access the fragment being read. For
instance, if we consider the previous example, the complete view may show all
fragments that have the same category in the context of all fragments. An
in-out-out-in subflow corresponds to a situation when the
reader stops reading, according to the current user activity, and chooses a new user
activity in which she continues to read.
Overall, the reading flow corresponds to the implementation of the in-out metaphor, where the level of disruption is minimal in in-in subflows, and maximal in in-out-out-in subflows. Note that to reduce the disruption of the reading
flow, all changes can be done using shortcuts, pressing a single key, and most menu
elements fade-out in the reading interface when the reader does not use the mouse so
that the screen only contains the text of the fragment.
4.2. User Interface Structure: Menus and Navigation
In the context of an edition, the LdoD Visual has three main
types of elements: selection menus, reading view, and global views.
4.2.1. User Activity Selection Menu
Once a context of an edition and one of its fragments is chosen, the main selection
menu allows users to choose between the available user activities, given the
context, and access the corresponding global view. The user activities are ordered
by semantic criteria: edition order, chronological order, text similarity,
categories/taxonomy, and heteronym.
As shown in Figure 3, each user activity is represented by a card with a title, a
button with the corresponding keyboard shortcut between straight parentheses that
will take the user to the activity, and an image that represents the information
visualization technique for that user activity, which can also be clicked to access
the user activity.
The image and button will be greyed out if the activity is unavailable - for example,
if the currently selected edition has no taxonomy, user activities that use this
semantic criteria category will not be available, displaying a message explaining
why it is not available in the title and its selection button is also
unavailable.
The same semantic criteria can be used with different information visualization
techniques. In Figure 3, we can observe that the top 2nd and
bottom 1st cards use different information visualization
techniques for chronological order, respectively, custom square map and timeline;
the top 3rd, bottom 3rd and 4th cards use the semantic criteria associated with the taxonomies, where
top and bottom 3rd use the same information visualization
technique, word cloud, the former in the context of the selected edition, and the
latter in the context of the selected fragment; while the bottom 4th uses a different visualization technique, network graph.
4.2.2. Reading View
In this view, the focus is on reading the text of a selected fragment, as represented
in Figure 4, where the fragment title and text are presented with a reading progress
bar. Since the text will be read on a screen, the used text font is a sans serif
font. The fragment title font and text font are the same ones used in one of the LdoD Archive reading interfaces.
Besides the progress reading bar, the reading view features some elements of
established browser book readers (see Figure 5). There are previous/next fragment
arrow buttons that retreat/advance through the selected edition fragments depending
on the currently selected edition. It also allows users to navigate between
fragments according to the user selected category or heteronym, using the yellow
arrows.
There are some detail features, such as smooth animation between fragment change and
smooth page scroll back to the top for the same matter. This smooth page scroll can
also be triggered by clicking on its matching button or pressing “T” on the
keyboard.
When the user stays inactive for more than 3 seconds, everything in this menu
disappears with the exception of the reading progress bar and the fragment's title
and text, as shown in Figure 4. Inactivity is timed when the user is not interacting
with any button, progress bar or previous/next fragment arrow buttons. It is then
possible to trigger everything to appear again by doing a mouse hover on any of
these elements of the menu. When the user scrolls down the page and the top buttons
overlap the text, there is also a transparency trigger. These features are activated
as an effort to maximize the user focus on reading the text and are another extra
step towards achieving our metaphor implementation.
In this menu, there is a user activity "under the hood" in the sense that it
resembles a user activity that does not break the reading flow: by clicking on the
matching button (“Highlight the most relevant words”), left button on top (shortcut
[R]), the 4 most relevant words are highlighted in blue, as presented in Figure 5.
This number was decided after extensive experimentation and fine tuning, trying to
find a balance between actual relevant content highlighting and not highlighting too
many words to the point it becomes irrelevant to do so. This option makes use of an
information visualization technique - text skimming - to encode a semantic criterion
- word relevance (more specifically, TF-IDF) -, even though it is not described as a
user activity in the sense that it can be applied independently of what is the
current user’s activity.
The reader can also see his/her own history of reading by pressing the button on the
right in the group of three buttons on top (shortcut [H]), as presented in Figure
6.
In this view, the reader is presented with an interactive timeline that displays the
fragments that have been read for the currently selected edition. Each history entry
is a fragment, displaying its title and a thumbnail that represents the information
visualization technique of the user activity through which that fragment was
reached. Besides, having a similar interaction to the user activity that makes use
of the timeline in the global view, the user can navigate back to a given fragment
that was previously read by clicking on its entry.
The buttons on top right, Figure 5, allow the user to see the instructions for this menu and
select another edition.
The two first buttons in the group of three buttons allow the user to access the
global view associated with the currently selected user activity (shortcut [A]) and
to navigate to the new user activity selection menu (shortcut [N]).
4.2.3. Global Views
The global views correspond to the out phase of the metaphor
and are a combination of semantic criteria with an information visualization
technique. In the following subsections we show how the global views are presented
in the LdoD Visual in the context of the various information
visualization techniques.
4.2.3.1. Network Graph
This information visualization technique consists of a network graph with hidden
edges, as shown in Figure 7. Each node is a circle that represents a fragment from
the currently selected edition. If the user clicks on a circle, she will be taken
back to the Reading Menu with the corresponding fragment of
that circle open to be read.
When accessing the view, the central orange circle represents the fragment in whose
context the user activity was started, and the purple circle represents the fragment
being currently read. This strategy was adopted so that the reading occurring during
a user activity is contextualized by the initially read fragment. The blue circles
represent the other fragments from the edition, and their relative distance to the
central circle expresses the encoded semantic criterion.
The user can put the mouse cursor over the circle so that a label is displayed,
showcasing the title and the value of similarity in percentage. Besides being able
to drag and zoom the network graph using the mouse pointer and mouse wheel, the user
can also use the green navigation buttons to drag, re-size, zoom in and out of the
network graph.
This information visualization technique is used to encode similarity by text and
taxonomy, which in the case of Figure 7 is by text.
4.2.3.2. Word Cloud
Word clouds are used to encode semantic criteria associated with the categories of a
certain fragment or taxonomy (group of categories) of the selected edition, as shown
in Figure 8. It displays each category using different colors and the font size
varies according to the number of fragments that belong to that category.
The Word Cloud, and its semantic criteria, can be applied in two cases: (1) the set
of categories of the taxonomy of an edition, which corresponds to the case presented
in Figure 8; (2) the categories of the current fragment. In both cases, by
selecting one of the categories, the user is redirected to a custom squares map that
highlights the fragments of the select category, as described in the next
subsection.
Heimerl et al. (
2014) developed a system that uses word clouds as its central
visualization method for interactive text analysis. Its results showed that even
though word clouds are aesthetically pleasing with its words disposed in different
angles, users do not find them very functional and easy to use. Thus, results showed
that participants preferred sequential layouts, where words are placed horizontally
without any kind of inclination or angles for aesthetic purposes. For this reason,
we decided to follow the same layout in the
LdoD Visual.
4.2.3.3. Custom Squares Map
This information visualization technique consists of squares linked by arrows, as
shown in Figure 9. It is similar to the network graph if we picture the squares as
nodes and the arrows as edges. Likewise, each square represents a fragment from the
currently selected edition.
This group of squares linked by arrows offers a global view of the currently selected
edition, and suggests a certain order derived from the arrows. This order depends on
the encoded criterion. Each square's color and highlighting also depends on the
encoded criterion.
Similarly to network graphs, there is a special color highlighting for each square.
The orange square corresponds to the fragment where the user activity was initiated,
and the purple square to the fragment currently being read.
This information visualization technique is used to encode the following semantic
criteria:
- Edition order.
Fragment squares are sorted by order of the currently selected edition. If
the user places the mouse cursor over a square, it will display a label with
the fragment title and its position in the edition, almost as if it were a
page number. This global view allows users to explore the order of the
fragments in the edition.
- Chronological order. Fragment squares are sorted by the date in which they
were written. Each square displays a two-digit number that
represents the year from that date - for example, a fragment from 1927 will
have its corresponding square with the number 27. Fragments without date
have their corresponding squares greyed out. If the user places the mouse
cursor over a square, a label will be displayed containing the fragment
title and its date. This global view allows users to explore the edition
ordered by date.
- Categories/taxonomy. Fragments are displayed exactly as if the semantic
criterion was the edition's order, with the addition of highlighting in
yellow the squares of the fragments from the selected category. If the user
places the mouse cursor over a square, a label will be displayed with the
fragment title and the categories to which it belongs, as shown in Figure 9.
As previously described, smaller yellow previous/next fragment arrow buttons
will appear under the black previous/next buttons in the reading
menu.
- Heteronym. It has
the same behaviour as the categories/taxonomy semantic criteria, but it
highlights in yellow the squares of the fragments that were assigned to a
specific heteronym. The yellow arrow buttons are used to move exclusively to
the previous/next fragment belonging to the
selected heteronym. The global view allows users to explore the fragments
assigned to a particular heteronym.
4.2.3.4. Timeline
This information visualization technique consists of an interactive timeline, in
which the user is presented with the timeline centered on the time window around the
currently selected fragment, as shown in Figure 10.
There is only one global view that uses this information visualization technique to
encode the chronological order as the semantic criterion, in order to explore the
fragments around the date of the current fragment. Each timeline entry represents a
fragment from the currently selected edition. Being consistent with the color scheme
for other visualizations, an entry will be orange if it represents the fragment that
was initially selected in the context of the user activity, and purple if it is the
fragment currently being read.
The user can navigate to a certain fragment by clicking on its entry.
5. Results and Discussion
The evaluation of the LdoD Visual was done by two types of
tests: usability tests, which assess the system’s usability, independently of the
its particular purpose; and, utility tests, which assess the quality of the system,
considering its purpose as a reading tool for the Book of
Disquiet.
5.1. Usability Tests
To assess how usable, efficient and satisfying the interaction with the LdoD Visual is, we have performed tests with 11 volunteer
users who have no expert knowledge on the LdoD Archive and on
the Book of Disquiet. The results of this type of test are
objective and quantitative.
Each usability test consisted of 4 stages:
- Introduce the user to the LdoD Visual's
concepts and briefly demonstrate the system in about 5 minutes;
- Let the user explore the system freely for about 5
minutes;
- Ask the user to execute 10 tasks;
- Ask the user to fill and submit a SUS (system
usability scale) questionnaire.
In the third stage, for each task, we counted the number of errors and time taken to
complete the task. The tasks assess the usability and intuition of the user
activities' interaction with the information visualization techniques in order to
explore and navigate through the fragments from the selected editions. For instance,
to select any fragment from an edition, select the user activity that makes use of
the squares map encoding the edition's order, choose the first fragment, highlight
the most relevant words and say out loud what are the most relevant words of the
selected fragment, which was, actually, task 9.
These tasks were always executed in the same order for each user, exactly from the
same menus.
Figure 11 presents the results from which we conclude that, on average, every task was
completed without exceeding the expected time and with basically no errors, with the
exception of the last task, which was more difficult. Task 10 was conceived to
confirm if users understood why a user activity was not available for certain
fragments and what they should do in order to select other fragments that supported
that previously unavailable user activity. This was the most complex task, which
explains why it took the most time out of all tasks, even though most users ended up
figuring out what they had to do.
For the fourth stage, the users filled and submitted a SUS (system usability scale)
questionnaire. This is a quick questionnaire that is commonly used to figure out if
any aspect was not considered by the tests. This questionnaire involved 10
statements for the user to answer how much he would agree with each one of them in a
scale from 1 (completely disagree) to 5 (completely agree):
- I think I would like to use the LdoD Visual
regularly.
- I think the LdoD Visual was unnecessarily
complex.
- I think the LdoD Visual was easy to use.
- I think I would need help from a person with
technical knowledge to use the LdoD
Visual.
- I think that all of the LdoD Visual
features were well integrated.
- I think the LdoD Visual shows a lot of
inconsistency.
- I think people will learn how to use the LdoD Visual
easily.
- I think the LdoD Visual was very confusing to
use.
- I felt confident while using the LdoD Visual
- I needed to learn a lot of new things before I could
use LdoD Visual.
The score of the SUS questionnaire is obtained by transforming the 1 to 5 scale in a
0 to 4 scale, and converting the values to be consistent, e.g., the odd-number
questions express a positive opinion and even-number questions express a negative
opinion. Figure 12 presents the results, where the questions' answer values are as given by the subjects, and the sums, averages and SUS final scores are calculated on the converted values.
Analyzing the SUS score, we can observe that the lowest scoring answer is number 4,
I think I would need help from a person with technical
knowledge to use LdoD Visual. This makes sense as the LdoD Visual is relatively complex to users that have no experience with
the Book of Disquiet and the LdoD
Archive, which is the case of the participants. Regarding the final score,
when a system scores a SUS score of 80.3 or more, it is considered to be in the top
10% of scores and the testing users are more likely to recommend it to others. This
means that if we are considering a SUS scale, the LdoD Visual
is a system with good usability, having a SUS score of 88.41.
Besides these 10 SUS questions, we also added questions in order to know about the
users. In terms of the characterization of the usability test participants: 9.1%
where younger than 18, 36.4% between 18-24, 45.5% between 25-34, and 9.1% between
45-54; and 9,1% have secondary education, 45.5% have a bachelor degree, and 45.5% a
master degree. Most users had already used other software to read books in the past
and most users had not read the Book of Disquiet. It is also
possible to observe that most users strongly agreed that they would have a good
reading experience using the LdoD Visual.
5.2. Utility Tests
The purpose of the utility tests is to obtain information about the utility and
quality of the experience while using the system. These tests are important in the
sense they allow us to assess subjective and qualitative aspects of the LdoD Visual, which are difficult to be measured
quantitatively.
The tests were done with 3 users that have expert knowledge about the Book of Disquiet and the LdoD Archive. The user of the first
case study is a postdoctoral researcher with a PhD thesis on the Book of Disquiet. This user has also worked, between 2012 and 2015, with
the XML encoding of the fragments for the LdoD Archive. The
user of the second case study is a PhD student from the PhD Program in Materialities
of Literature, which is a doctoral program that addresses the material and
technological mediations of literary practices. One of its research fields is
digital humanities. This user has a BA and an MA in graphic design. The user of the
third case study is another PhD student from the same PhD Program, who has a BA in graphic design, and an MA in Semiotics.
This user is interested in the aspects of usability
of the LdoD Archive since her PhD project involves creative
practices through the situated use of the archive. She is one of LdoD Archive's most regular users and has been organizing a plan of
activities to teach LdoD Archive's users how to use it and
fully explore its various functionalities, including the creation of virtual
editions.
The protocol for this type of testing is more flexible in comparison to the
usability tests' protocol. The users were contacted by email, introduced to the LdoD Visual metaphor, and encouraged to explore the system
before we had our actual meeting to do the case study.
During the case study the users were encouraged to use the LdoD
Visual while “thinking out loud”. We asked them several questions regarding
the real utility, potential and reading experience that they thought the LdoD Visual provides. We also asked them about the
differences they felt as more important, in comparison to other tools and electronic
readers available, as well as comparing the reading experience of the LdoD Visual to the other reading features available in the
LdoD Archive.
Considering the feedback received from these case studies, we can conclude that the
reception of the LdoD Visual is positive, both in terms of
concept, execution and utility.
The users liked the approach on how to read a book. They thought that it is a good
match for the fragmentary nature of the Book of Disquiet. The
attempt to output a visual representation of the editions and fragments brings a new
perspective on how to read and explore the Book of Disquiet
and use the LdoD Archive, highlighting its relevance for
the field of digital humanities.
The second user also described her experience with user activities in a way that
suggested that the LdoD Visual's metaphor is successfully
implemented. She stated: “The real reading customization strength comes from the
activities around the currently selected fragment. If the reader stops and tries,
for example, to continue her reading path with a user activity that is related with
reading similar fragments by text to the currently selected fragment, there is
already a reflexive question that is specific to the reader and the navigation
method centered around the selected fragment becomes the medium for the user to be
able to read around an idea that is interesting to her. In LdoD
Visual, the reader is able to navigate through the fragments according to
what is desired, having a visual reference of the type of navigation and exploration
that is being performed in the Book of Disquiet.”
The users also suggested that the LdoD Visual is relatively
complex and that there should exist more ways of welcoming non-expert users who are
not aware of the problem and structure of the Book of
Disquiet and the the LdoD Archive, such as "selling
the concept" both as a ludic and research tool, and showcasing the different uses
and possibilities in videos scattered around the application.
The users also acknowledged the potential of the LdoD Visual
concepts such that they could be developed and expanded to explore other
alternatives of interaction with the Book of Disquiet.
We could observe that each user spent time with different parts and user activities
of the LdoD Visual. Our analysis on this is that the LdoD Visual seems to provide different affordances for
different users, and that the objective of customizing the exploration and reading
experience of theBook of Disquiet was met.
We can also conclude that the LdoD Visual is a tool that
mainly scripts two different reading behaviours: the possibility of navigating
through the Book of Disquiet with a focus on the pleasurable
assisted user-driven reading experience, geared towards a continuous and immersive
flow; and the possibility of navigating with a humanities researcher mindset,
finding patterns, correlations and textual data insights, according to a
discontinuous and interruptive logic. Both of these reading practices are possible
resorting to the same mechanisms and representational strategies, i.e., the user
activities' information visualization techniques, even though the first type of
reading practice is the main focus of the LdoD Visual.
6. Conclusion
The Book of Disquiet by Fernando Pessoa is an unfinished book
that can be read in any order. The modular nature of this fragmentary literary work
brings endless possibilities on how it can be traversed.The goal of this work was to
design and implement a web-based application to read The Book of
Disquiet, giving the reader an active role while reading and browsing
through the book's different fragments, defining user activities that make use of
information visualization techniques to encode different semantic criteria.
In order to achieve this, we researched on the use of information visualization
techniques applied to textual data and literature, and on the concept and tools for
e-books and web-based reading. Afterwards, we defined the in-out-in metaphor that was used as the guiding concept for whether or not
a certain architecture or feature choice met the objectives and served the overall
purpose of this work.
According to this metaphor, we defined the user activity concept, which establishes a
perspective on how to read the book. A user activity is a combination of an
information visualization technique with a semantic criterion, where the latter is
used to classify the fragments and the former to visualize them according to the
classification. The solution architecture considers three types of user interfaces:
a user interface in which the user activity is selected; a global view in which the
fragments are visualized according to the chosen user activity; and a reading view
to support the reading and navigation between fragments, according to the chosen
semantic criteria.
The final system was tested for usability and utility, and it got positive feedback
from both non-expert and expert users of the Book of Disquiet
and the LdoD Archive. The usability tests have shown
that it is possible to complete the defined tasks in the expected time and without
errors, and the utility tests also confirmed that the proposed metaphor meets the
experts’ vision of the book and how to explore its multiple paths of reading.
As a web-based application for exploring the process of reading, the
LdoD Visual contains several features that correspond to what
the
LdoD Archive describes as “the simulation principle,”
that is, the ability to provide interactors with a reflexive engagement with the
textual environment [
Portela and Magalhães 2020]. Through a recursive process of
going into the text and coming out of the text, readers are able to explore the
Book of Disquiet as a particular kind of reading experience
and they are also able to see how their acts of engaging with the text become
registered in the visualizations. The implemented visualization techniques thus
bring together the modularization of the text and the modularization of the reading
of the text.
The LdoD Visual is a tool that has the potential to grow and
become an improved version of what it already accomplishes. One major improvement
would be the possibility of saving the reading history state to a user account,
preferably in the LdoD Archive. This would be a logical
extension since the process of reading a book can take days, weeks or months.
Another improvement would be developing a feature to generate new virtual editions
according to specific reading paths. This would explore the LdoD
Archive goal of fostering the construction of new virtual editions of the
Book of Disquiet. In its reflexive graphical
representation of reading activities, this visual reader contributes to the
experimental textual rationale of the LdoD Archive.
We believe the In-Out-In metaphor adopted in this project can also be used as a
conceptual design tool for modelling reading processes in digital media. The design
of interfaces for complex archives of literary materials based on this model would
have twofold benefits: limiting hypertextual disorientation, on one hand, and
supporting a meaningful transition between human and machine reading processes, on
the other, for both general and expert readers. In this light, the model of the
reading flow embodied in the LdoD Visual could also be seen
as a prototype of a specifically digital paratextual apparatus for the smooth and
optimal integration of close reading and distant reading modes through multiple
visualization techniques.
NOTE: In order to keep track of the project’s evolution, the source code is publicly
available in a GitHub repository
https://github.com/socialsoftware/edition.
Acknowledgements
The authors wish to thank the anonymous reviewers of this article for their
invaluable comments and suggestions.
This work was supported by national funds through FCT,
Fundação para a Ciência e a Tecnologia, under projects UIDB/50021/2020 and UIDB/00759/2020.
Works Cited
Alexander and Gleicher 2016 Alexander, Eric, and Michael Gleicher (2016). “Task-driven comparison of topic
models. IEEE Transactions on Visualization and Computer Graphics.” 22(1),
320–329.
Baron 2015 Baron, Naomi (2015). Words Onscreen: The Fate of Reading in a
Digital World. Oxford: Oxford University Press.
Bernstein 2011 Bernstein, Mark (2011). “Can We Talk about Spatial Hypertext?” HT
'11: Proceedings of the 22nd ACM conference on Hypertext and hypermedia,
June 2011, 103–112. https://doi.org/10.1145/1995966.1995983
Bolter 2001 Bolter, Jay David (2001). Writing Space: The Computer, Hypertext,
and the History of Writing. London: Routledge [2nd ed.].
Brath and Ebad 2014 Brath, Richard, and Ebad Banissi (2014). “Using font attributes in knowledge maps and
information retrieval.” Proceedings of the First Workshop on
Knowledge Maps and Information Retrieval (KMIR 2014)co-located with International Conference on Digital Libraries (DL 2014),
23–30.
Brath and Ebad 2015 Brath, Richard, and Ebad Banissi (2015). “Using text in visualizations for
micro/macro readings.” 4th Workshop on Visual Text Analytics. ACM, 2015.
Cheema et al. 2016 Cheema, Muhammad Faisal, Stefan Jänicke, Gerik Scheuermann (2016). “AnnotateVis:
Combining Traditional Close Reading with Visual Text Analysis.” Workshop on
Visualization for the Digital Humanities, IEEE VIS 2016, Baltimore, Maryland, USA,
October 24th, 2016. URL:
http://www.informatik.uni-leipzig.de/~stjaenicke/annotatevis.pdf
Coelho 1982 Coelho, Jacinto do Prado, ed. 1982. Fernando Pessoa, Livro do Desassossego por Bernardo Soares.
Collection and transcription by Maria Aliete Galhoz and Teresa Sobral Cunha. Lisboa: Ática, 1982. 2 vols.
Drucker 2014 Drucker, Johanna (2014). Graphesis: Visual Forms of Knowledge
Production, Cambridge, MA: Harvard University Press.
Eaves et al. 1996-2021 Eaves, Morris, Robert N. Essick, and Joseph Viscomi, eds. (1996-2021). The William Blake Archive. Chapel Hill, NC: University of
North Carolina at Chapel Hill.
Hayles 2012 Hayles, N. Katherine (2012). How We Think: Digital Media and
Contemporary Technogenesis. Chicago: The University of Chicago Press.
Heimerl et al. 2014 Heimerl, Florian; Steffen Lohmann; Simon Lange; and Thomas Ertl (2014). “Word Cloud
Explorer: Text Analytics based on Word Clouds.” 47th Hawaii
International Conference on System Sciences, Jan 2014, 1833–1842. DOI:
10.1109/HICSS.2014.231
Herzog 2016 Herzog, Werner, dir. (2016). Lo and Behold: Reveries of the Connected
World(2016). New York: Magnolia Pictures.
Film.
Jänicke et al. 2015 Jänicke, Stefan, Greta Franzini, Muhammad Faisal Cheema, and Gerik Scheuermann
(2015). “On Close and Distant Reading in Digital Humanities: A Survey and Future
Challenges.” Proceedings of the Eurographics Conference on
Visualization (EuroVis) Cagliari, Italy, 25–29 May 2015. http://www.informatik.uni-leipzig.de/~stjaenicke/Survey.pdf
Jänicke et al. 2017 Jänicke, Stefan, Greta Franzini, Muhammad Faisal Cheema, and Gerik Scheuermann
(2017). “Visual Text Analysis in Digital Humanities.” Computer
Graphics Forum, 36(6), 226-250. https://doi.org/10.1111/cgf.12873.
Koolen et al. 2012 Koolen, Coorina, Alex Garnet, and Raymond G. Siemens (2012). “Electronic Environments
for Reading: An Annotated Bibliography of Pertinent Hardware and Software.” Scholarly and Research Communication, 3(4), 1-62.
Landow 2006 Landow, George P. (2006). Hypertext 3.0: Critical Theory and New
Media in an Era of Globalization. Baltimore, MD: The Johns Hopkins
University Press.
Mangen et al. 2016 Mangen, Anne, and Adriaan van der Weel (2016). “The evolution of reading in the age
of digitisation: an integrative framework for reading research.” Literacy, 50(3), 116-124.
https://onlinelibrary.wiley.com/doi/abs/10.1111/lit.12086
Mangen et al. 2019 Mangen, Anne, Gérard Olivier, and Jean-Luc Velay (2019). “Comparing comprehension of
a long text read in print book and on Kindle: Where in the text and when in the
story?” Frontiers in psychology: cognitive science. 15
February 2019 https://doi.org/10.3389/fpsyg.2019.00038
Manovich 2011 Manovich, Lev (2011). “What is visualisation?”, Visual
Studies, 26(1), 36-49. DOI:10.1080/1472586X.2011.548488
Marshall et al. 1995 Marshall, Catherine C., and Frank M. Shipman (1995). “Spatial hypertext: designing
for change.” Communications of the ACM, 38(8), 88-97. August
1995 https://doi.org/10.1145/208344.208350
Mohammad 2012 Mohammad, Saif M. (2012). “From once upon a time to happily ever after: Tracking
emotions in mail and books.” Decision Support Systems, 53(4),
730-741. DOI: 10.1016/j.dss.2012.05.030
Moulthrop 1991 Moulthrop, Stuart (1991). “You Say You Want a Revolution? Hypertext and the Laws of
Media”. Postmodern Culture, 1.3. DOI: https://doi.org/10.1353/pmc.1991.0019
Moulthrop and Grigar 2017 Moulthrop, Stuart, and Dene Grigar (2017). Traversals: The Use of
Preservation for Early Electronic Writing. Cambridge, MA: The MIT
Press.
Murray 2012 Murray, Janet H. (2012). Inventing the Medium: Principles of Interaction Design as
Cultural Practice. Cambridge, MA: The MIT Press.
Nelson 1965 Nelson, Theodor H. (1965). “A File Structure for the
Complex, the Changing, and the Indeterminate.” Association for Computing Machinery: Proceedings of
the 20th National Conference, 84–100.
Nelson 1995 Nelson, Theodor H. (1995). “The Heart of Connection:
Hypermedia Unified by Transclusion. ”Communications of the ACM 38(8), 31–33.
Nelson 2003 Nelson, Theodor H. (2003). “Computer Lib/Dream Machines”
(1974). The New Media Reader, ed. Noah Wardrip-Fruin and Nick Montfort. Cambridge, MA:
MIT Press. 303-338.
Pizzaro 2010 Pizarro, Jerónimo, ed. 2010. Fernando Pessoa, Livro do Desasocego. Edição Crítica das Obras de Fernando Pessoa, Vol. XII (Tomos I e II). Lisboa: Imprensa Nacional-Casa da Moeda.
Portela 2017 Portela, Manuel (2017).
“The Book of
Disquiet Archive as a
Collaborative Textual Environment: From Digital Archive to Digital
Simulator.” The
Writing Platform: Digital Knowledge for Writers. Brisbane: Queensland
University of Technology. Web. URL: https://thewritingplatform.com/2017/07/book-disquiet-archive-collaborative-textual-environment-digital-archive-digital-simulator/
Portela 2019 Portela, Manuel (2019). “The LdoD Archive as a creative
textual environment and a model of literary performativity.” Computational Creativity Meets Digital Literary Studies, Dagstuhl Reports,
Edited by Tarek Richard Besold, Pablo Gervás, Evelyn Gius, and Sarah Schulz, 9(4),
87–106 [p. 98].
Portela 2022 Portela, Manuel. “From
Meta-Editing to Virtual Editing: The LdoD Archive as a Computer-Assisted
Editorial Space.” Approaches to Teaching Pessoa's The Book of Disquiet. Eds. Paulo de Medeiros and
Jerónimo Pizarro. New York: Modern Language Association. [forthcoming].
Portela and Magalhães 2020 Portela, Manuel, and Cecília Magalhães (2020). “The Book of Disquiet
Digital Archive as a Role-Playing Experiment.”
Attention à la Marche! Mind The Gap! Thinking Electronic
Literature in a Digital Culture. Eds. Bertrand Gervais and Sophie Marcotte.
Montréal: Les Presses de l'Écureuil. 307-325.
Portela and Rito Silva 2015 Portela, Manuel, and António Rito Silva (2015). “A model for a virtual LdoD.” Digital Scholarship in the Humanities, 30(3), 354-370.
http://dx.doi.org/10.1093/llc/fqu004
Portela and Rito Silva 2017 Portela, Manuel, and Rito Silva,
António, eds. (2017). Arquivo LdoD: Arquivo
Digital Colaborativo do Livro do Desassossego. Coimbra: Centro de Literatura
Portuguesa da Universidade de Coimbra. URL: https://ldod.uc.pt/
Rito Silva and Portela 2015 Rito Silva, António, and Manuel Portela (2015). “TEI4LdoD: Textual Encoding and
Social Editing in Web 2.0 Environments.” Journal of the Text
Encoding Initiative, 8. URL : http://journals.openedition.org/jtei/1171.
DOI : 10.4000/jtei.1171
Singhal 2001 Singhal, Amit (2001). “Modern Information Retrieval: A Brief Overview.” Bulletin of the IEEE Computer Society Technical Committee on Data
Engineering 24 (4), 35–43.
Sobral 2008 Sobral Cunha, Teresa, ed. 2008. Fernando Pessoa, Livro do Desassossego por Vicente Guedes, Bernardo Soares. Lisboa: Relógio d’Água. [First ed. 1990-91]
Solís and Ali 2011 Solís, Carlos, and Nour Ali (2011). “An Experience Using a Spatial Hypertext Wiki.”
HT '11: Proceedings of the 22nd ACM conference on Hypertext
and hypermedia, June 2011, 133-142. DOI: 10.1145/1995966.1995986
Theng and Thimbleby 1998 Theng, Yin Leng, and Harold Thimbleby (1998). “Addressing Design and Usability Issues
in Hypertext and on the World Wide Web by Re-Examining the ‘Lost in Hyperspace’
Problem.” Journal of Universal Computer Science, 4(11),
839-855.
Theng et al. 1996 Theng, Yin Leng, Matthew Jones, Harold Thimbleby (1996). “‘Lost in hyperspace’:
Psychological problem or bad design?” Proceedings 1st Asia-Pacific
Conference on HCI, 387-396.
Vilaplana and Pérez-Montoro 2017 Vilaplana, Jaume Nualart, and Mario Pérez-Montoro (2017). “Diggersdiaries: Using text
analysis to support exploration and reading in a large document collection.” Eurographics Conference on Visualization (EuroVis), Posters Track
(2017), A. Puig Puig and T. Isenberg, Eds. 9-11. DOI:
10.2312/eurp.20171156
Viégas et al. 2009 Viégas, Fernanda B., Martin Wattenberg, and Jonathan Feinberg (2009). “Participatory
visualization with wordle.” IEEE Transactions on Visualization and
Computer Graphics, 15, 1137–1144. DOI: 10.1109/TVCG.2009.1
Zenith 2012 Zenith, Richard, ed. 2012. Fernando Pessoa, Livro do Desassossego. Lisboa: Assírio & Alvim. [First ed. 1998]