AbstractThis article introduces a tool for the visualization of indexations of articles about digital literature works. Resulting from an international project gathering various analytical perspectives on a limited body of works, this tool constitutes a proposition in the debate on the preservation of works. It allows one to visualize and compare the different points of view and the parts of the work that have been analyzed. This tool considers the work as an entity within a Spinozist ontology. The tool is based on the procedural model of digital works developed by Philippe Bootz. These two theoretical models (Spinozist ontological approach to the work and procedural model) are presented in this article, before we tackle the issue of indexation from a technical point of view.
1. IntroductionThe ELO directory, along with the ELMCIP and CELL projects, develop huge databases gathering data on digital literature works and theoretical articles. Methods to visualize the data they contain are required, in order to ensure an efficient access to this wealth of data. Such methods are closely related to both the choices made with regards to the indexation of these data and the visualization objective. Now, while we can content ourselves with structural information on works, leaving them open to interpretation, a documentary database of close readings is interesting only if their content can be visualized. The importance of such visualization is obvious when it comes to research: knowledge of the works is contained in essays, not in the works themselves. And yet most of the visualization methods used in artistic analyses are based on the visualization of parts or characteristics of the works. This is the case of Media Visualization1 or direct visualization, which consists in creating visual representations from media or media parts that are initially visual (images, videos, etc.). This method is particularly used in the research on cinema and art history. The Distant Reading2 method is suggested in the quantitative study of global literature and is also based on the collection and analysis of a great number of texts. Therefore, moving away from these methods and using scientific articles as a documentary database implies a reflection on the —semantic— indexation and visualization of these articles.
This is what justified the "Augmented textualities in context" research project which spanned over a year. It gathered researchers from several teams, with a view to create a greatly varied body of close readings. The participants were: Philippe Bootz, Marida Di Crosta, María Inés Laitano, Estrella Rojas and Alexandra Saemmer (from the Écritures et Hypermédiations Numériques team of the Laboratoire Paragraphe), Sandy Baldwin, James B. Bishop and Dibyadyuti Roy (from the Center for Literary Computing), Claudia Kozac (from Ludión), and Arnaud Regnauld (from the Laboratoire Transferts Critiques et Dynamiques des Savoirs).
When we noticed that articles tackle works according to varied research objectives and methodologies, it became obvious that the indexation and visualization of these articles should not attenuate these differences, but on the contrary show them, and therefore focus on the issue of the points of view. What points of view does an article express on a work, and what parts of the work do these points of view focus on precisely? Without talking about a metric distance, can one find elements that bring them closer, or on the contrary, that takes them away from one another despite their resemblance? This kind of tool would be most useful for research.
In a previous project, Philippe Bootz and Samuel Szoniecky developed a tool for the indexation of points of view, named archipoenum3. Therefore, we have at our disposal a theoretical approach capable of processing these points of view. The present project aims to determine whether adopting this approach is relevant with regards to our research issue. To do so, we have developed, as part of this project, a visualization prototype called Cross-reading and an indexation method suited to the articles.
2. Methodology of the ProjectInitially, we selected a restricted body of seven works4, each project participant committing to analyze at least two works, so as to obtain several analyses per work. Everyone was free to choose their method and analysis objectives, as well as the implemented concepts.
3. Theoretical Foundations
3.1. An Approach to the Work Based on the Spinozist Ontology
3.1.1. The Main Concepts of the Spinozist OntologyThe idea of the nature of the digital works analyzed in the close readings, on which Cross-reading is based, derives from Deleuze's perspective on the Spinozist ontology6. Note that in this ontology, individuals are defined by three dimensions: their essence, their extensive parts and the singular relations linking the essence to the extensive parts. The extensive parts of individuals consist of the set of physical components that are somehow related to their essence. Their number is infinite and indeterminate. They are mortal, death consisting in the cancellation of the relations between these physical elements and the essence. In this way, the extensive parts constitute a quite broader vision of the body. The essence is an individual's personal agency, it pertains to the spirit, to the mind. It is eternal yet can only be expressed through extensive parts. Deleuze extends this ontology from the realm of humans to every physical body. For instance, in his course on Spinoza7 , he mentions the extensive parts of the wave.
Spinoza defines three kinds of knowledge an individual can have of another: the inadequate knowledge resulting from a contact (which he calls collision) between some of these two individuals' extensive parts; they are only related to the extensive parts. He puts the affects at this level. The secondary knowledge is that of relations. Every scientific knowledge falls within this category. The operating mechanisms of the extensive parts, which science explains, do indeed very well describe the singular relations between these parts and the essence. Knowledge of the essences is intuitive.
3.1.2. Transposition onto the Digital WorkAt first sight, one may find it shocking to consider a work as an individual or human subject, since the work apparently does not share all the characteristics of a human being. What about the human's free will, for instance? What about the human's decision power? By doing so, are we not personifying the work, which would simply amount to using the rhetorical figure of allegory instead of a scientific reasoning?
From a strictly ethnocentric point of view, only humans are individuals. As they approach the world from their point of view and according to a certain relationship, the physical components coming into contact become their extensive parts. This collection process does not require very strong relations between these new extensive parts and the essence: the relations between this glass in which I am drinking and my essence are weaker than those between this glass and my arm. Philippe Bootz uses the term "parergonic" to describe these extensive parts having a weak relation with the essence. They indeed share with the parergon discussed by Derrida8 this specificity of being both internal and external, a kind of boundary between the extensive parts and external materiality. Humans in contact with the work thus turn the components of the work into extensive parts of themselves, according to the relations determined by the relationship they have with the work (creation, reading, analysis, restoration, adaptation, indexation...) Then, they act on other human individuals through the consequences of this capture, namely through the creation of new material elements that will be captured as extensive parts by these other individuals. In this way, the program created by the author will produce a phenomenon on the screen, which will be captured by the reader. The researcher's production, for example a close reading, may influence the reader's interpretation of the work. Therefore, it is clear that the work, as the parergonic body of several individuals, constitutes a media through which these individuals act upon one another. When one examines the relation between a given individual and the work, it is strictly equivalent to considering that the set of human influences acting through the work on this individual can be modeled through the unique personal agency the work is ascribed with, which no human individual can handle.
The work begets the work only because it possesses this personal agency, or because human individuals act on other human individuals through the relation that they have with this work, which amounts to the same thing. Therefore, regarding the work as a Spinozist individual is a purely intellectual operation, similar to a change in reference point in the field of physics.
The essence of the work now defined, what about the relations? The relations linking the extensive parts to the essence, in this model, are described by the set of concepts that can be applied to the work. These relations are therefore accessible to knowledge, to every form of knowledge. For example, when the movement of a word allows for a set of singular interpretations, it is according to the concepts and mechanisms of a semiotic and rhetorical model. These concepts give an indication on the relations between this movement and the essence of the work from the specific semio-rhetorical perspective. However, this particular movement may also allow to set the work in the heritage of a previous literary or artistic movement. Artistic movements are categories induced by a specific division of the world, they are concepts. The name of the movement and filiation process are therefore indications on the relations that this extensive part of the work bears with its essence, from the perspective of art history. As one can notice, relations in this model are only comprehended through points of view on the world, the number of which is indefinite. Besides, these points of view evolve, appear and disappear over time, and above all, are specific to a given culture. In this spatiotemporal process of the evolution of points of view, it is as if the relations transformed because of the evolution of the work's essence, which results from the diversification of the individuals acting through the work (today an author, tomorrow a critic, later a custodian when the author has disappeared...), an evolution that is unpredictable yet based on reasoned or emotional considerations, hence similar to a free will in a way.
3.2. The Procedural ModelThe procedural model describes the extensive parts of the work. During a contact between a human subject and the work, the former enters some kind of relationship with the work. The human subject acts through the work, and as such, both participates and is subjected to the latter's personal agency. Therefore, through their activity, subjects are captured by the work as much as they capture it. The roles played by human individuals in these activities thus constitute extensive parts of the work. As well, documents such as design documents, authors' words, readers' reactions... are directly related to this activity and enter an indirect relation with the essence of the work. They partly result from the action that the work performs on the subjects and allow the work to indirectly act in the world. In this way, they also constitute as many extensive parts of the work, although they are linked to its essence by weak relations. They are part of the parergonic body of the work, contrary to the program (the source9) and the media result (the transitoire observable), which are produced during the execution of the work and in very close relation with the essence of the work. The procedural model calls "digital body" the material extensive parts that are linked to the essence of the work through strong relations. It distinguishes between three types of extensive parts: the roles, the digital body of the work (which includes the source and the transitoires observables) and its parergonic body, which is made up of various documents. These extensive parts are interrelated according to the following diagram, in which the digital body is in gray and the parergonic body in yellow:
The source and transitoire observable are material elements that can be recorded or manipulated by a (peripheral) device such as a camera or a scanner. When human subjects approach them, they necessarily process them on a semantic mode, retrieve meaning of any kind, and turn them into a set of signs respectively named texte-auteur (author-text) and texte-à-voir10 (text-to-see). The documents dealing with material parts of the work, most often what is displayed on the screen —a texte-à-voir— or characteristics of the program —an texte-auteur— refer to these signs, or more exactly to specific parts of these signs.
3.3. Ontology of the Document in ArchipoenumIn the procedural model, researchers perform a meta-reading activity and their articles constitute secondary discourses. Conversely, one can consider that, through this research activity, the material parts of the works examined by a researcher become his/her extensive parts, according to a specific relation to his/her essence determining the researcher's point of view on these extensive parts. The areas of the signs: texte-auteur and texte-à-voir are then components of this relation, because they result from a semiotic decision largely based on the relations that the researcher has with the world. It is through their interpretation that the point of view is expressed.
The document model developed by archipoenum draws on the previous considerations and uses a property of the procedural model: a given human subject can successively play several roles and a researcher can shift from a perspective to another. Each point of view is expressed in a specific secondary discourse within the article. The data model of archipoenum calls them sections. Each section develops a single point of view, via concepts indicating the specific relations that the operator has with an area of signs according to this point of view. They constitute the ontology11 of the operator according to that point of view. The section is the basic unit that is subjected to indexation. Depending on the granularity of its delimitation, indexation will be more or less accurate.
In this approach, indexing documents consists in interpreting them from the procedural point of view and according to a Spinozist ontology; in projecting them onto the same interpretive mold. In this way, projecting heterogeneous documents according to a same point of view allows for their comparison.
4. The Cross-Reading PrototypeEvery visualization software demands that one defines a model of the data to visualize. The visualization software of Cross-reading is based on a simplified version of the data model used by the indexation tool of archipoenum. It proceeds from a description of the work and a description of the scientific articles.
4.1. The Data Model
The essence of the work being uncontrollable, it is only detected by the title of the work. The work can be developed in several versions (adaptations on various platforms, updates, performed versions, etc.) which, in compliance with the data model of archipoenum, constitute relations (in the sense defined by Spinoza) allowing for the categorization of sets of extensive parts, one per version. Most often, a given article only deals with one version of the work. For each version, the procedural model allows for the distinction of several kinds of extensive parts: digital body, parergonic body and roles.
In the procedural model, the different close readings visualized by the tool consist of secondary discourses. As we have just mentioned, they do not directly deal with the extensive parts, but the sets of signs that the analyst has built on these extensive parts: texte-auteur or texte-à-voir depending on the object of the article12. Therefore, every document only deals with specific parts of an texte-auteur or texte-à-voir, which the data model identifies as "textual objects".
The model of document which is used to process the articles is based on the fundamental concept of section, as introduced in archipoenum. Cross-reading does not aim at visualizing all the information contained in an article, but only the information related to the sources and transitoires observables. We have adjusted the granularity of the sections, in order to get only one concept per section. The section of a document is then defined by a unity of textual object and a concept that ascribes it with a unity of point of view on this textual object. Some parts of the articles do not deal with a specific textual object but rather with a context or referent outside the work.
The data model of Cross-reading is then described by the following Entity/Association model, in which the author entity corresponds to three different activities: the author role of the procedural model, the researcher as the author of the document (in a situation of meta-reading in the procedural model) and the author of the ontology concepts. These three activities can be carried out —or not— by the same person.
Note that while the current prototype only mentions the sections relating to sources or transitoires observables, the model allows for the integration of any section related to the components accessible through analysis, namely relations 1 to 7 of the procedural model.
The data model provides for the possibility of clarifying relations of synonymy and homonymy between concepts. This possibility favors the translation of an ontology into another and allows for a qualitative comparison of ontologies. Sometimes, determined by its definition, a concept may be ascribed with a certain label in an ontology and another label in another ontology. The opposite can also occur. These cases are likely to happen when ontologies from different disciplines are applied to a same work in different articles. The definition of the synonymy and homonymy of concepts is certainly anything but trivial, but we have not found the time to tackle this issue in Cross-reading. As a result, these relations have not been used.
4.2 IndexationOnce this data model had been established, the project participants indexed their articles by identifying the sections according to their beginning and end expressed in the form (page, line). They have specified the name of the concept that was used in the considered section, as well as the author of this concept, and indicated the title of the work to which the section referred.
Therefore, this indexation does not compel the indexer to subscribe to the procedural model in any way: the indexation of an article does not require a specific vocabulary of the model.
Although the relations of homonymy and synonymy have not been used, we have asked the researchers to add a glossary of the concepts used, with a view to implement these relations at a later stage.
Here is an extract from Arnaud Regnauld's indexation of his article "Interrupting D: Patchwork Girl's Syncopated Body", which deals with S. Jackson's Patchwork Girl:
P.1 – L13 to L31: concept syncope: undecidability (Jean-Luc Nancy); work Patchwork girlThe projection onto the procedural model was performed by the person in charge of the development of the visualization tool, M. Inés Laitano. For every indexed section, she has identified the textual object which was referred to by the section and the corresponding stimulus (source or transitoire observable). Obviously, the identification of textual objects is important and should be standardized to avoid any confusion. In this prototype, we have not dealt with this aspect, the names of the objects were generally defined within the section by the researchers themselves and no confusion was reported during the validation sessions.
P.1 – L35 to L46: concept signature (Jacques Derrida); work Patchwork girl
P.2 – L1 to L3: concept signature (Jacques Derrida); work Patchwork girl
P.2 – L4 to L7: concept spectrality (Jacques Derrida) work Patchwork girl
P.2 – L8 to L13: concept phallogocentrism (Jacques Derrida) work Patchwork girl
P.2 – L13 to L15: concept supplement (Jacques Derrida); work Patchwork girl
4.3. The Visualization of DataThe next step consisted in translating the entities of the data model that were identified through indexation into graphic primitives. It seemed to us that the network visualization was best suited to the ontology of Figure 2. Each entity thus produces a network node and every relation between entities gives rise to an arc. The type of entity is represented in the network by the color of the node. The cardinality indicated on the ontology relations ("one to many" for example) becomes clear in the network, since it is possible to count the arcs between two nodes. Figure 3 below represents the analysis of A. Regnauld's Patchwork girl. It shows the network corresponding to the previous indexation. This visualization shows that A. Regnauld only focused on the transitoire observable, not on the source. It also shows that he analyzed several specific textual objects, identified by the blue nodes, and used four concepts borrowed from Derrida and one from Nancy in his analysis.
The attributes of ontology that are specific to each kind of entity are not directly visible in this network, so as not to impede its readability, but they can be consulted for every node by rolling the mouse cursor over this node. A pop-up window then gives the detailed information in the data model of the considered node:
Even in its minimal form (Figure 3), the network is packed with information. As soon as one adds other researchers' analyses, it becomes illegible. This is why we have scripted it, so as to gradually display information according to temporal sequences depending on the network user's visualization objectives.
The initial view only shows the body of analyzed works through a visualization of the nodes related to the titles of the works (Figure 5, upper left). By clicking on the "work title" node, the user displays the nodes of the versions of the work that have been analyzed, its authors, and the titles of the documents that analyze them (Figure 5, upper right). In this view, clicking on the node of a version results in the display of the nodes corresponding to the elements of the digital body (source or transitoire observable) with which the article has dealt, as well as the analyzed textual objects related to these elements (Figure 5, lower left). In the second step, it is also possible to click on the node of the document. This results in the display of the indexed sections and the concepts developed in these sections, as well as the authors of these concepts (Figure 5, lower right). In this way, one can develop the network over time, until it is completely visualized (Figure 3).
The user has ample opportunity to visualize any partial view of the network, according to her visualizaton objective. The tool helps her isolate the sub-network corresponding to a certain section, concept, or textual object, by clicking on the node of this section, concept, or textual object. For example, this manipulation allows for the individual analysis of each and every concept tackled in an analytical document. Figure 6 shows the visualization of the sole section 0 of Regnauld's document. The initial scripting (Figure 5) only aims to have the reader gradually discover the structure and extent of the information. Besides, the network can be spatially rearranged through drag and drop in every view.
4.3.1. Comparisons of AnalysesCross-reading allows to comprehend the various approaches to a work at a glance, as well as the concepts used in these approaches. When two analyses are carried out from the same perspective, it is possible to draw a kind of comparison. This is what happened in this project: Philippe Bootz and Alexandra Saemmer have both adopted a semiotic approach to analyze Alexandra Saemmer's Tramway. Although Ph. Bootz's analysis focuses more on the source than the transitoire observable, while A. Saemmer only worked on the latter, it appears that both of them evoke the concept of imitation to talk about this version of the work. The implementation of the relations of homonymy and synonymy would have proven useful to draw a conclusion from this apparent identity.
5. Conclusion and ProspectsThe preliminary study conducted as part of this project has shown that the procedural model could prove useful to visualize the information about a work which was contained in close readings and, more generally, in research articles. This tool is not a substitute for these analyses, it only provides a medium to visualize relations between the said analyses and to carry out searches in the body of works or that of analyses.
Several points still require development and standardization. Moreover, indexation is manual and one may wonder whether this tool can be adapted to an automatic semantic indexation. It is not easy to answer this question, because one has to make sure that the considered indexation mode preserves the points of view and that the concepts do not get drowned in a statistical analysis of the vocabulary that is used.
In any case, such visualization tool can be used to compare points of view, complete critical analyses, help Ph. D. students to find leads that have been but little pursued, to facilitate the trans-cultural analysis of the circulation of concepts, or even, through a diachronic use, to track the evolutions of notions.
Even in its current manual version, this tool could be usefully interfaced with the ELMCIP database, provided that it is completed with an indexation field according to the suggested model. It is indeed possible to change program and visualization type in order to use this base, yet the data model of Cross-reading should be preserved.
Manovich, Lev. 2010. "What is visualization?" Poetess Archive Journal 2 (1). http://journals.tdl.org/paj/index.php/paj/article/view/19. back
Moretti, Franco. 2000. "Conjectures on World Literature." New Left Review II (1). http://newleftreview.org/II/1/franco-moretti-conjectures-on-world-literature. back
Bootz, Philippe, Samuel Szoniecky, and Abderrahim Bargaoui. 2013. "Entité / Identité. Un Outil D'indexation Des Documents Relatifs à La Poésie Numérique." Revista Cibertextualidades, no. 5. back
In the White Dakness by Rainer Srasser; Patchwork girl by Shelley Jackson; Inside. A Journal of Dreams by Andy Campell; Works2 by Iván Marino; Migraciones by Leonardo Solaas and ppg256 by Nick Montfort. back
Available on the website: http://dddl.eu/outil-indexation/. back
Bootz, Philippe, and Samuel Szoniecky. 2013. "Vers Une Ontologie Du Domaine de La Poésie Numérique." Revista Cibertextualidades, no. 5. http://bdigital.ufp.pt/bitstream/10284/3865/1/cibertextualidades_05_bootz-szoniecky.pdf. back
Deleuze, Gilles. 2001. Spinoza, éternité et immortalité. Cours donné en 1981 à l'Université Paris 8. Paris: Gallimard. CD. back
Derrida, Jacques. 1978. La Vérité en peinture. Paris : Flammarion. back
The source consists of the material parts that are manipulated through the author's activity. Therefore, the source is neither the program, nor the data. The program does not appear in its binary form, it is the source program, as computer specialists call it, or in other words the form that can be understood and manipulated by a human being, which appears in the language or software used by the author, together with the data and media as processed by the author. back
By using Klinkenberg's definition of the sign, this transformation constitutes a semiotic decision, in which the part of the source involved in a texte-auteur constitutes its stimulus, just like the part of the transitoire observable involved in the texte-à-voir is the stimulus of the latter. back
In the computer sense of the word. back
We have not visualized the data from the articles that concern physical components other than the works' sources and transitoires observables. back