FiCT-O: Modelling Fictional Characters in Detective Fiction from the 19th to the 20th Century

Abstract

This paper proposes a formal descriptive model for understanding the evolution of characters in detective fiction from the 19th to the 20th century, using methodologies and technologies from the Semantic Web. The integration of Digital Humanities within the theory of comparative literature opens new paths of study that allow for a digital approach to the understanding of intertextuality through close reading techniques and ontological modelling. In this research area, the variety of possible textual relationships, the levels of analysis required to classify these connections, and the inherently referential nature of certain literary genres demand a structured taxonomy. This taxonomy should account for stylistic elements, narrative structures, and cultural recursiveness that are unique to literary texts. The detective figure, central to modern literature, provides an ideal lens for examining narrative intertextuality across the 19th and 20th centuries. The analysis concentrates on character traits and narrative functions, addressing various methods of rewriting within the evolving cultural and creative context of authorship. Through a comparative examination of a representative sample of detective fiction from the period under scrutiny, the research identifies mechanisms of (meta)narrative recurrence, transformation, and reworking within the canon. The outcome is a formal model for describing narrative structures and techniques, with a specific focus on character development, aimed at uncovering patterns of continuity and variation in diegetic content over time and across different works, adaptable to analogous cases of traditional reworking and narrative fluidity.

1. Introduction1

the character is a product of combinations: the combination

is relatively stable (denoted by the recurrence of the semes)

and more or less complex (involving more or less congruent,

more or less contradictory figures); this complexity determines

the character’s “personality,” which is just as much a combination

as the odor of a dish or the bouquet of a wine

(Barthes [1970] 1974)

In the ever-expanding network of disciplines within Digital Humanities (hereon DH), the analysis of literary phenomena must be reconsidered in light of four interrelated dimensions: the research purpose, representing the qualitative or epistemic core of the literary inquiry; the scope, or the quantitative dimension that defines the corpus under examination; the computer-aided method employed, which is not merely instrumental but constitutive of the research outcome; and finally, the inherent complexity of the literary datum itself.

Focusing on the interplay between the first two dimensions, the variations in their balance have generated a wide range of disciplinary approaches, which are often treated as conflicting and rather should be analyzed in terms of complementarity. With a certain extent of reduction, we identify, on the one hand, methodologies that tend to emphasize the granularity and intensive capacity of the analysis of single works or a small set of works—thus enhancing the expressiveness of the data; on the other hand, disciplinary approaches that instead prioritize the computability of large corpora, and thus employ large-scale, “bird’s-eye” approaches (Horstmann et al. 2024). In the first case, the analysis includes any form of qualitative, philological, or contextual interpretation applied to individual texts (or small corpora) commonly linked to the rather ubiquitous notion of close reading (Wilkens 2012; Smith 2016; Kestemont and Herman 2019). In the second case, it designates a range of disciplines that, as Ciotti defines, “encompass quantitative and large-scale data-driven analyses of literary phenomena, conducted through sophisticated computational methods” (Ciotti 2021), variously connected to the concept of distant reading, as introduced by Franco Moretti in 2000 (Moretti 2000, 2013a).

A widely held view is that the rationale underlying the latter studies—regardless of how consequential their results may be—requires that literary data be “flattened” through processes such as tokenization, topic modelling or text mining (Moretti 2013b; Schöch 2017; Schöch et al. 2022), sacrificing the semantic complexity generated by interpretation in favor of computable description (Piper 2019). From this standpoint, distant reading has come under scrutiny for a range of reasons: scholars have variously questioned its limited hermeneutic depth (Ascari 2014; Drucker 2017), its lack of specificity in supporting close textual analysis (Tally 2007; Bode 2017), and its failure to establish a cohesive theoretical framework (McCarty 2008; Ciotti 2021). By contrast, it bears restating that close reading, even when understood as a more strictly defined and digitally grounded methodological framework (e.g., Digital Scholarly Editing, XML/TEI encoding, and RDF modelling), often remains constrained by corpus scope and semantic granularity (Jockers 2013; Krautter 2024), thus requiring complex metadata stratification, which perpetuates the long-standing issue of overlapping, widely acknowledged within the field (Pierazzo 2015; Italia 2020; Gallucci 2023).

Following Piper (2015) and Underwood (2017, 2019), this paper advocates for a more integrative view, in which distant and close reading are not conceived as opposing or mutually exclusive methods, but rather as complementary modalities within a broader methodological continuum, defined by a fluid variation in the scope of analysis. This continuum aims to achieve a synthetic integration of computability and expressiveness, aligning large-scale formal modelling with the interpretive depth characteristic of traditional textual analysis. As Underwood notes, “distant reading is simply a new scale of description, no more in conflict with close reading than an anatomical diagram of your hand would be with the chemical reactions going on inside your cells” (Underwood 2019).

The methodological potential of DH lies precisely in the capacity of “zooming-in” and “zooming-out” between distant and close readings—what Martin Mueller termed scalable reading (Mueller 2012, 2014), prerequisite that considers both qualitative and quantitative factors to tackle the complexity of humanistic data (Weitin and Werber 2017), embedding different levels of text interpretation and formalizing such concepts to make them interoperable and computable, so as to ensure a “scalable workflow” (Horstmann 2020).

Building on these preliminary considerations—which have addressed the first two dimensions introduced at the outset—it is now possible to clearly define the third dimension of the method: a computer-aided process that enhances the semantic and expressive potential of computational approaches to literary criticism (Ciotti 2016; Tomasi et al. 2021), while simultaneously acknowledging how the integration of observational data can expand the informational capacity of digital technologies (Dennett 1990; Sanfilippo et al. 2021).

The methodological approach adopted in this study is rooted in the dual imperative of rendering literary texts computable—that is, expressed through a formally defined, syntactically and semantically unambiguous representation (Piper 2019)—and of making explicit the semantic and hermeneutic complexity that such texts inherently convey. This requires a modelling process capable of declaring the classes and core concepts that define the relevant domain of knowledge, as well as the properties and relationships among its constituent entities.

These principles are foundational to the Semantic Web, which provides a technical and epistemological infrastructure for structuring data through ontologies understood as “formal, explicit specifications of a shared conceptualization” (Gruber 1993). At its core lies the DIKW hierarchy (Data, Information, Knowledge, Wisdom), which guides the transformation of raw data into structured knowledge, a crucial design when dealing with literary data: to this extent, data do preserve their original literary significance when embedded within an informational context (Lotman [1970] 1980; Eco 1979), and become knowledge when subjected to interpretive processes (Ricoeur 1983–1985). The architecture utilizes RDF for graph-based representation, OWL (Web Ontology Language) for formalizing logical hierarchies and relationships, and Description Logic—known as the T-Box/A-Box structure—to enable machine reasoning within the domain. This approach allows even the complexities of literary data to be modelled as computational objects that possess semantic expressiveness.

However, there remains a persistent misalignment between methodology and data, leading to a conceptual gap. This ultimately raises a critical question: how can one make sense of what is semantically opaque? As Ciotti highlights, “data in literary studies do not precede formal modelling; on the contrary, they are the product of modelling” (Ciotti 2016, p. 33). In computational narratology, the trade-off between expressiveness and computability centers on two core challenges: modelling the inherently plural and layered nature of interpretative acts and addressing the ontological status of fictional characters. This includes examining their metaphysical grounding, referential validity, and how they can be formally represented within logical systems.

Addressing the former, we advocate anchoring semantic formalism in narratological theory, which abstracts narrative structures (e.g., time, voice, focalization, and actants) into computable elements. The goal is to provide a clear semantics capable of analyzing character functions and roles within plot structure (Greimas, Propp, Todorov), the mechanisms underlying narrative construction (Chatman, Genette, Uspenskij, Segre), and the dynamics of textual reception across the “triangle” of text, authorship, and interpretation (Barthes, Foucault, Eco). While remaining aware of the intrinsic limitations of any such formalization and acknowledging that “formal modelling is inside the hermeneutic process and that we are expected to modify and adapt it, ad infinitum” (Ciotti 2016, p. 35), such modelling fosters scalable interpretability across corpora.

The second issue—the ontological nature of fictional characters—invites philosophical reflection, particularly when engaging with ontologies. While this topic opens up a longstanding and complex debate that exceeds the scope of this article, it remains crucial to lay down a foundational theoretical framework to inform the modelling approach employed in this study. As widely acknowledged in both analytic metaphysics and narratology (Friend 2007; Kroon and Voltolini 2018; Damiano et al. 2019; Hastings and Schulz 2019; García-Carpintero 2020), fictional characters exhibit a dual ontological status: they seem to both exist and not exist (Walton 1990; Priest 2013; Rouillé 2023) while “the reader is not confused about the real existence of the fictional entity” (Ingarden 1973). This paradoxical nature has led many scholars to conceive fictional entities as bundles of properties attributed to them as if they were real individuals, thereby enabling rigorous semantic analysis without entailing any ontological commitments. This approach resonates strongly with Pretense Theory (Walton 1990; Thomasson 1999; Kroon and Voltolini 2018; Tomazzoli and Favazzo 2024), which frames discussions about fictional characters as a form of a structured imaginative practice, where characters are described and referred to as if they were real, rather than making literal statements about real entities. Under this interpretative lens, characters serve as representational constructs defined by specific traits and internal consistencies, a role that highlights their pivotal status within narratological inquiry (Eder et al. 2010). As Marchese famously stated, “character remains the real problem” (Marchese 1983), encapsulating the persistent complexity of this notion. Semiotic approaches further emphasize the multifaceted identity of fictional characters, shaped not only by their narrative roles and interrelationships with other entities, but also by their development across various literary traditions, genres, and thematic systems.

Within the domain of the Semantic Web, various efforts have been undertaken to formalize the concept of the fictional character, aiming to address the theoretical and representational challenges it poses (Zöllner-Weber and Witt 2006; Damiano and Lieto 2013; Bartalesi et al. 2016; Hastings and Schulz 2019; Schöch et al. 2022; Sanfilippo et al. 2023).2 Particularly noteworthy is the recent GOLEM ontology (Pannach et al. 2024), which offers one of the most innovative and up-to-date contributions to the field of semantic modelling applied to literary studies: a high-granularity domain-specific ontology for semantic annotation, narrative structures, textual variants, and authorial interpretations—bridging computational formalism with interpretive depth.

A computational methodology of this kind may contribute to the identification of recurring patterns across diachronic corpora. The goal is not to flatten meaning but to work across various levels of abstraction. By doing this, it clarifies terminological and semantic distinctions, fostering a productive integration between close and distant reading practices.

With this framework established, this paper explores the core Research Question (RQ): How has the representation of the detective figure evolved in detective fiction from the 19th to the 20th century?

The first dimension of purpose consists of adopting a bottom-up approach to semantic modelling, based on the specific techniques of literary analysis. In terms of scope, this study conducts a close reading of Arthur Conan Doyle’s works to identify the core ontological features of the character Sherlock Holmes. It then evaluates how these features are transformed into two 20th-century rewritings: La verità sul caso D. by Fruttero & Lucentini, and Le Loup chantant de Forest Gate by the Ouvroir de Littérature Policière Potentielle group (hereon OuLiPoPo). Since detective fiction is fundamentally characterized by a “clear narrative structure” (Petronio 1985), it serves as an ideal context for multi-layered analysis and ontological formalization. Therefore, the character of Sherlock Holmes is analyzed both as a narrative function—an “external” construct shaped by the author’s design within the plot—and as an individual with “internal” traits like those of real-world entities. This dual analytical perspective supports the development of FiCT-O (Fictional Character Ontology), the ontology proposed in this study. It is designed to model the layered nature of fictional entities through domain-specific formalism and cross-ontology alignment mechanisms.3

This paper is organized as follows: Section 2 provides a detailed overview of the selected case studies, explaining the rationale behind their selection and outlining the methodology adopted for their analysis. Emphasis is placed on the construction of a knowledge base, defined here as a formally structured and semantically enriched repository of information that is encoded in machine-readable formats to support automated querying, inference and knowledge discovery. Section 3 introduces a series of Competency Questions (hereon CQs) formulated using SPARQL, the standard query language for retrieving and manipulating data in knowledge bases grounded in the RDF model. These questions have been designed to test whether the developed ontology can accurately and consistently represent the intended concepts. Section 4 addresses interpretative outcomes and limitations, concluding with a discussion of existing gaps and forthcoming challenges.

2. Materials and Methods

2.1. Methods

In recent years, an increasing number of scholars have explored how Applied Ontology (Munn and Smith 2008) and Semantic Web technologies can model literary knowledge domains by structuring and operationalizing data (Moretti 2013b) while preserving the semantic depth intrinsic to literary texts. Whether pursuing qualitative or quantitative purpose, and regardless of the breadth of the analytical scope (see Section 1), these technologies offer a robust formal framework for knowledge representation, allowing for the explicit definition of a domain’s representational primitives (Gruber 2009)—i.e., the foundational elements and conceptual building blocks that define the structure of knowledge within a specific field.

Ontological modelling formalizes key elements clearly and consistently, ensuring they are both machine-readable and computationally actionable. This framework is particularly important in literary domains, where meaning is not fixed but negotiated through the interplay of the author’s intentional act (Wimsatt and Beardsley 1946; Dennett 1990) and the hermeneutic context of reception (Barthes 1967; Foucault 1969; Eco 1979). In narrative texts, meaning is layered and contingent, shaped by authors, readers, and the broader cultural, temporal, and intertextual contexts in which texts circulate. As a result, computational tools that address this domain must be capable of declaring descriptive properties that account for such complexity. Semantic Web-based ontological modelling offers a methodologically sound approach to formalizing the dynamic construction, negotiation, and transmission of narrative meaning across interpretive contexts, while preserving the ambiguity and semantic richness that characterize literary discourse (Tomazzoli and Favazzo 2024). As outlined in the introduction (Section 1), this study employs Semantic Web formalism to model the layered nature of the interpretive act that constitutes the fictional horizon of narrative texts. This process involves mapping essential narrative elements—such as characters, themes, roles, and points of view—using shared and machine-processable vocabularies. Furthermore, it seeks to provide Semantic Web methodologies with a comprehensive theoretical foundation grounded not only in taxonomic structure but also in a semantic-interpretive network informed by narratological and semiotic theory. Rooted in Russian Formalism, French Structuralism, and New Criticism, these frameworks define the intentional and representational components of narrative by means of formalized categories that identify the essential features “that make a given text a narrative text” (Marchese 1983, p. 7).

Against this background, the present study adopts specific narratological theories both to formalize the structural constituents of fictional narratives and to inform the close reading of selected case studies. Semantic Web technologies prove particularly effective in this context, as they make explicit how the selection of an analytical model constitutes an interpretive and epistemological act—one that enables the extraction and systematization of additional semantic layers into a coherent and computationally actionable structure. From early Russian Formalism to the neo-structuralist approaches of the late 20th century, narratological inquiry has progressively shifted from focusing solely on the content of the story to a more refined analysis of the mechanisms through which it is constructed and conveyed—mechanisms that, particularly through the seminal work of Gérard Genette, have been conceptualized as formal strategies governing the relationship between narrated events and their narrative mediation.

To maintain analytical clarity while avoiding conceptual overlap, this study adopts a two-level model that distinguishes between: (1) diegetic components, which broadly aligned with the notion of fabula, referring to the world of the story as constructed through events, characters, temporalities, and settings; and (2) narrative construction (or sujet), which denotes the discursive organization and articulation of the story, including elements such as narrative voice, event sequencing, point of view, and focalization (Tomaševskij [1925] 1965; Barthes 1971; Genette [1972] 1976; Bal [1980] 1990).

Special attention is given to Genette’s structural categories:

(1)

Narrative instance or the narrator’s ontological and functional position about the story, classified along the axes of extradiegetic/intradiegetic and heterodiegetic/homodiegetic;

(2)

Focalization, i.e., the filtering perspective of narrative information, ranging from zero to internal and external focalization;

(3)

Modes of discourse, encompassing the stylistic and structural forms used to convey speech, thought, or action, including narrated discourse, direct and indirect speech, and free indirect discourse.

As a preliminary step in the analysis, and in response to a major methodological gap in current Semantic Web applications in the literature, the feasibility of mapping such formal features is tested within a genre characterized by high structural regularity and semiotic coherence—namely detective fiction—whose codified structure and persistent narrative architecture make it an ideal case study for modelling fundamental narrative categories. Defined as “a type of popular literature in which a crime is introduced and investigated, and the culprit is revealed” (Matković 2018, p. 446), detective fiction exhibits a formulaic structure that typically unfolds through an ordered sequence of narrative phases (Priestman 2003). From Cawelti (1976) to Rzepka (2005), scholars have emphasized its reliance on a codified progression—from the presentation of a crime to its resolution—and on recurring roles such as detective, victim, criminal, and ancillary figures. This structural rigidity, governed by genre conventions and interpretive expectations, makes detective fiction a narrative laboratory for examining plot mechanics and functional character roles (Todorov 1966; Eco and Sebeok 1983).

Within such a highly codified narrative form, the relationships between character ontology and plot function appear closely interdependent, as characters tend to be defined less by psychological depth than by the structural roles they occupy in the narrative economy. These roles, functioning as formal positions in the diegetic system, activate the semiotic apparatus of the fictional world and contribute to its internal coherence. As a result, the act of narration becomes inseparable from the construction of fictional entities, which serve as both products and agents of a regulated narrative sequence.

The scope of the research focused on a selection of works from the detective fiction tradition, centering on Arthur Conan Doyle’s early and late Sherlock Holmes stories—A Study in Scarlet, The Hound of the Baskervilles, and The Final Problem (Doyle 2015a, 2015b, 2018). This selection allows for a comparative mapping of narrative configurations, particularly the relationship between the narrator and the story. In all three texts, the narration is consistently homodiegetic, mediated by Dr. Watson. However, the internal architecture varies, with shifts occurring between intra- and extradiegetic layers, as well as between internal focalization (Watson’s viewpoint) and zero focalization (the narrator’s superior knowledge). These variations elucidate how narrative voice and information control are strategically deployed within a rigid framework.

The corpus serves as an initial dataset modelling two key dimensions (see Section 2.2): (i) each text and its associated bibliographic and contextual metadata; and (ii) its diegetic features, especially character construction and narrative structuring devices.

An opening remark involves the multilingual and transnational nature of the corpus, which includes texts originally written in English (A Study in Scarlet, The Hound of the Baskervilles, The Final Problem), French (Le Loup chantant de Forest Gate), and Italian (La verità sul caso D.). This distribution is significant: while detective fiction originated in United States, it quickly spread and evolved within other literary traditions, particularly in France, England, and finally in Italy (Pistelli 2006, pp. 3–5). Focusing on the first three linguistic and cultural contexts allows for an examination of the genre’s historical dissemination and the diverse ways its conventions have been reinterpreted across linguistic and cultural boundaries. In this context, the inclusion of two 20th-century rewritings—Le Loup chantant de Forest Gate by the OuLiPoPo group (OULIPOPO 1975) and La verità sul caso D. by Fruttero et al. (1989)—serves a crucial heuristic function, as both texts engage in metafictional plays and were selected for their capacity to invoke and subvert the narrative conventions of detective fiction, ultimately challenging the coherence of the fictional world they inherit.

This selection enables a dual inquiry: first, into how a shared narrative modality is transformed across distinct literary and cultural contexts; second, into how Holmes’s identity is reshaped in derivative texts. By tracing which narrative and semantic properties are preserved or modified, the analysis brings to light the semiotic elasticity of fictional characters across intertextual frameworks, with the selected works offering insights into the broader dynamics between authorship—represented by the original creator of a fictional entity (Conan Doyle)—and the intertextual processes that redefine that character within new narrative structures. Whether through constrained experimentation (Le Loup chantant de Forest Gate) or literary pastiche (La verità sul caso D.), both texts exemplify how ontological identity and narratological function become areas of negotiation in postmodern fiction.

As anticipated in Section 1, a second major challenge in applying Semantic Web technologies to literary studies concerns the ontological status of fictional characters. While formalist and structuralist traditions have refined the analysis of narrative logic, they often overlook the character as a psychological or ontological unit. Classical narratology has largely conceptualized characters not as autonomous agents but as structural functions within the plot, as in Propp’s morphology or Greimas’s actantial model, where characters are essentially reduced to narrative “masks” that serve causal progression and teleological structure. In this theoretical framework, characters operate as functional nodes within the syntagmatic structure of the récit, serving as stereotypical placeholders whose meaning lies in either facilitating or obstructing narrative development. As Marchese observes, “the character remains the most complex element in narratology” (Marchese 1983), precisely because it resists complete reduction to formalized roles—a tension that has reverberated into computational narratology, where such structural reductionism continues to inform several ontological modelling.

Indeed, this reductive tendency is evident in many literary-oriented ontologies (e.g., Damiano and Lieto 2013; Bartalesi et al. 2016), which often emphasize narrative processes while relegating characters to interchangeable slots among settings, events, and objects. These frameworks tend to underrepresent the psychological, stylistic, and semiotic complexity that defines character identity within a diegetic world. In contrast, the present analysis underscores the necessity of treating fictional characters not merely as narrative functions but as semantically complex entities with identifiable traits that can be understood both internally and externally. Drawing on insights from cognitive science, which analyzes mental states via their causal and functional interdependencies, fictional characters are here approached as complex semantic constructs whose actions reflect interconnected traits.

Rather than engaging with debates on the ontological status of fictional beings as “real” entities, this approach adopts the perspective of Pretense Theory of fictional discourse (Thomasson 1999; Walton 1990; Byrne 2001), which interprets statements about fictional entities as part of a structured “game of make-believe”. From this standpoint, fictional entities are conceived as bundles of properties invoked in imaginative practices, allowing for their modelling within formal ontologies without committing to their actual existence. Consequently, fictional characters can be represented as a specific subtype of Information Content Entities (hereon ICEs), as defined by Smith and Ceusters (2015) and Hastings and Schulz (2019). ICEs are generically dependent on material bearers—such as physical or digital texts—and are about something: they refer to entities, whether they are real or imagined. This as-if-aboutness enables the modelling of fictional characters as analyzable, intelligible constructs within semantic systems. By integrating their dual nature—as both structural functions and semantically coherent referents—fictional characters may be formalized as epistemically robust entities. This modelling strategy preserves their expressive and interpretative richness of literary characterization (Eder et al. 2010), while ensuring formal computability within Semantic Web frameworks, without requiring ontological commitment to their existence.

To further articulate the roles and narrative functions of fictional characters, particularly detectives, this analysis incorporates Greimas’s actantial model (Greimas 1966, 1979), which classifies characters according to six narrative functions: subject, object, sender, receiver, helper, and opponent. These functional positions, though schematically defined, are often fluid and dynamically redistributed throughout the narrative, offering a valuable analytical framework for mapping how character roles evolve in response to plot development. Particular attention is devoted to the interaction between the detective and other actants, enabling a functional reading of the narrative system that highlights structural dependencies and oppositional dynamics. To complement this structuralist perspective, the model is further enriched through the contributions of Marchese (1983), whose theoretical insights allow for the formalization of individualizing traits such as psychological disposition, physical appearance, and ideological stance. By integrating these two dimensions—the structural and the expressive—fictional characters can thus be conceptualized as semantically rich and context-sensitive entities, simultaneously fulfilling functional roles within the narrative architecture and manifesting complex semantic constructs within formal ontological representations.

Given the ontological representations must remain anchored to the textual fabric from which fictional characters emerge, a key methodological assumption underpinning this approach is that modelling must be grounded in a close textual analysis. Each character is analyzed through textual segments that convey salient identity-defining traits, with a particular attention to the detective figure as a paradigmatic agent of genre-specific function. These “minimum textual units”—ranging from short expressions to complex sentences—are identified and categorized based on their adherence to genre conventions and their inclusion of at least one distinctive trait relevant to character development.

The third modelling requirement, which complements structure and context, involves the identification of recurring and meaningful traits associated with the central character, Sherlock Holmes. Based on the previously outlined framework, the application of semantic technologies aims to formalize key dimensions of character representation: the evolution of traits across different texts, the recurrence of symbolic or thematic patterns, the psychological articulation of attributes, and the influence of narrative perspective on the reader’s perception.

Uno studio in rosso, the Italian translation of A Study in Scarlet (1887), was chosen as a prototypical case study. This choice ensured methodological consistency across the multilingual corpus by distinguishing source and translation versions in the annotated dataset. The use of translations—specifically, the Italian versions based on Doyle’s English originals—reflects a critical awareness that translation is not a neutral transfer of content. Instead, it reshapes narrative elements such as voice, tone, and perspective (Hermans 2007; Venuti 2017; Munday et al. 2022). These intermediations are essential to how fictional characters are received and reinterpreted, and they must be acknowledged as part of their interpretive ecosystem.

The textual analysis thus focuses on the figure of Sherlock Holmes across both intra- and extra-diegetic dimensions: (1) his individuated traits, treated as if real and referential; (2) his narrative function within the fabula and his interaction with other characters; (3) his representation through the narrator’s voice, which shapes epistemic access and perception. This threefold perspective corresponds to three complementary dimensions of Sherlock Holmes’s characterization. First, all descriptive content concerning Holmes is filtered through the narrative voice of Dr. Watson, who simultaneously functions as both narrator and character within the diegesis (3). Second, within the narrative structure, Holmes assumes the role of subject-actant, while Watson performs the helper function, instrumental to the resolution of the plot (2). Finally, Holmes emerges as a richly individuated figure—investigator and flat mate at 221B Baker Street—defined by deductive brilliance, behavioral idiosyncrasies, and a distinctive philosophical outlook (1). This narratological framework guided the formalization of Holmes’s character through the identification of 70 annotated textual instances—each corresponding to a sentence or clause—resulting in the extraction of 104 distinct properties, which form the foundation of the conceptual model (see Section 2.2).

Following the initial character modelling based on A study in Scarlet, the dataset was further enriched through the inclusion of two additional canonical texts by Arthur Conan Doyle: The Hound of the Baskervilles (1902), selected for its greater psychological and narrative complexity, and The Final Problem (1893), which marks the initially intended conclusion of Holmes’s literary arc. This phase brought the dataset to 125 annotated instances, from which 179 recurring properties were extracted. In a subsequent step, the corpus was then extended to include the two apocryphal works previously analyzed, La verità sul caso D. and Le Loup chantant de Forest Gate, to compare character traits from Doyle’s canon to those reimagined in distinct cultural and literary contexts. This allowed this study to reconceptualize intertextuality as a transhistorical reconfiguration of character traits, tracing Holmes’s evolution across traditions and systems of reception. An additional 43 instances were annotated from the apocryphal texts, yielding 67 new properties.

As a direct outcome of the preceding analytical process, the development of a set of CQs laid the groundwork for semantic modelling activities aligned with the FAIR principles (Findable, Accessible, Interoperable, Reusable4). To ensure interoperability and semantic alignment, the model’s development was informed by a survey of existing models in the cultural heritage domain (see Section 2.2). To express semantic complexity not addressed by foundational ontologies, it was deemed necessary to define custom classes and properties. Before converting the entities into a structured knowledge base, all were aligned with external semantic resources, including Wikidata,5 Getty AAT6 and DBpedia,7 to ensure accurate reconciliation and semantic enrichment. The primary outcome of this research is the development of FiCT-O, a formal ontology designed to enhance the computational analysis of recurring narrative structures, specifically using detective fiction as a key example. The final dataset was converted into an RDF knowledge base in Turtle format, modelled according to FiCT-O, and then queried using SPARQL. This process enabled both the validation of the ontology’s internal consistency and a data-driven investigation into the evolving configuration of the detective figure across both canonical and apocryphal texts within the genre.

2.2. The Model

The proposal model previously outlined (Section 1 and Section 2.1) aims to address three formalization requirements: (i) to document the literary works and their contextual information; (ii) to explore the diegetic dimension associated to each literary work (e.g., narrative structures and techniques, characters, and plots); and (iii) to emphasize the main character and its personal traits. To fulfil these requirements, the formal model incorporates established ontologies from the GLAM (Galleries, Libraries, Archives, and Museums) sector, along with custom classes and properties tailored specifically for narratological aspects. The integration of existing schemas, vocabularies, and ontologies serves the primary purpose already identified (Section 2.1): to ensure the reliability and interoperability of the model while aligning closely with cultural heritage standards. Furthermore, the inclusion of narratological theories provides scientific rigor to the model and situates it within the relevant domain.

Proceeding from the first requirement—namely, the representation of literary works and their contextual information—the modelling process adopts the conceptual structure defined by WEMI (Work, Expression, Manifestation, Item), as illustrated in Figure 1. Originally introduced in the FRBR (Functional Requirements for Bibliographic Records8) framework and further refined in the IFLA Library Reference Model (LRM9), this model provides a hierarchical approach to describing bibliographic entities.

For this research, only the first two levels of the WEMI structure—Work and Expression—are selected, along with the semantic relationship connecting them (lrmoo:F1_Work, lrmoo:F2_Expression, and lrmoo:R3_is_realised_in). This choice reflects the specific focus on the intellectual content of literary texts, as opposed to their material or publication-related characteristics (which would involve Manifestation and Item). Furthermore, the model incorporates the logic of narrative creation, emphasizing types of evolution within the literary genre through crm:E55_Type, lrmoo:F27_Work_Creation, and lrmoo:F28_Expression_Creation.

We reuse a set of unconstrained properties from the RDA (Resource Description and Access10) standard and incorporated the CIDOC-CRM (Conceptual Reference Model11) ontology to enhance the metadata associated with key contextual dimensions. These dimensions include: agents (rdau:P60447 “has creator”, rdau:P60385 “has translator”, crm:E39_Actor, lrmoo:F55_Collective_Agent); language (crm:P72_has_language, crm:E56_Language); and temporal scope (rdau:P60527 “has date of resource”, rdau:P61020 “has related timespan of resource”, crm:E52_Time-Span).

In order to illustrate the functioning of the ontology in modelling textual data according to the structure presented in Figure 1, the case of A Study in Scarlet offers a concrete example. The title, rdfs:label “A Study in Scarlet”, is assigned to an instance of lrmoo:F1_Work, which is in turn realized through an instance of lrmoo:F2_Expression. This expression, labelled “Uno studio in rosso”, was produced in 2015, with its temporal metadata captured via the property rda:P61020 “has related time span of resource”. The act of creating this expression is modelled as an instance of lrmoo:F28_Expression_Creation, whose type (crm:P2_type) is defined as a “translation” and reconciled with the Getty AAT through the URI http://vocab.getty.edu/page/aat/300027389 (accessed on 19 August 2025). The language of the expression is specified as “Italian”, using the property crm:P72_has_language. The translator, then, is modelled using the property rdau:P60385 “has translator”, and corresponds to the individual “Giancarlo Carlotti”, further identified through his VIAF authority record (owl:sameAs: http://viaf.org/viaf/71171353 (accessed on 19 August 2025)). Finally, the original author of the work is linked through rdau_P60447 “has creator” to the instance “arthur_doyle” (crm:E39_Actor), labelled “Arthur Conan Doyle” and reconciled with his VIAF identity owl:sameAs: https://viaf.org/en/viaf/65283845 (accessed on 19 August 2025).

For the second requirement, established narratological theories are integrated with the reuse of additional ontologies beyond those previously mentioned. These include PROV-O (Provenance Ontology12), CiTO (Citation Typing Ontology13), and the Ultralite version of DOLCE (Descriptive Ontology for Linguistic and Cognitive Engineering14). In line with this multi-ontological integration, the model—as visualized in Figure 2—conceptualizes narrative structures as a distinct intellectual creation (ficto:Narrative), aligned with both dul:Narrative and crm:E28_Conceptual_Object and linked to an agent through the property crm:P94_has_created. Each narrative is composed of one or more narrative segments (ficto:NarrativeSegment), connected via dul:hasPart, and directly associated with the literary resource from which they are extracted (prov:value). These segments are identified based on their linguistic and functional characteristics, which reflect established techniques for organizing narrative discourse (Brémond 1974; Genette [1972] 1976).

Building on the bibliographic framework presented in the first part of the model, we can now illustrate the semantic representation of narrative and discursive levels, outlined in Figure 2, through a concrete example drawn once again from A study in Scarlet. The entire narrative content embodied in this expression (lrmoo:F2_Expression) is instantiated as a ficto:Narrative (“narrative_1”) (Presutti and Gangemi 2016), and is further structured into 70 distinct segments (ficto:NarrativeSegment), using the dul:hasPart property relation. Each segment is associated, via prov:value, with a textual excerpt relevant to the semantic description of the narrative discourse or to the characterization of key entities, particularly Sherlock Holmes. For instance, the segment narr_seg_expr1.3 corresponds to the sentence “No, non è uno che si sbottona molto, anche se può essere abbastanza espansivo quando gli salta il ticchio”, which conveys a psychological trait of Holmes—his reserved disposition combined with occasional bursts of sociability—formalized as a ficto:Attribute of the frbroo:F38_Character instance representing Holmes. This attribute is linked to the narrative segment through ficto:hasAttribute, thus enabling the integration of textual fragments into the structured semantic modelling of narrative functions and character profiles. As discussed in Section 2.1, according to Genette’s classification, each narrative segment can assume one of four possible forms of speech (ficto:hasFormOfSpeech), depending on the degree of the narrator’s involvement in the story. Each segment is produced by a narrative instance (ficto:hasNarrativeInstance), which is typically represented by a narrator (ficto:Narrator). The narrator’s position is defined by its relationship to the narrative and the diegetic levels (ficto:NarrativePosition, ficto:NarrativeLevel). Within this framework, the narrator can be classified as either intradiegetic or extradiegetic, and as homodiegetic or heterodiegetic.

An important aspect of narrative representation is the point of view, which is expressed through a type of focalization (ficto:Focalization). This determines the cognitive perspective from which the story is told. Following the tradition established by Marchese (1983) and Bal ([1980] 1990), we adopt an inclusive view that treats focalization as a formalized type of point of view, thereby avoiding excessive terminological fragmentation and enabling a coherent ontological representation of narrative perspective. With this goal in mind, the property ficto:hasPointOfView is used to link the narrator to the focalization strategy employed in the narrative, whether it is internal, external, or variable, while model also accounting for cases in which the narrator coincides with a character within the storyworld. In such instances, the narrative agent is classified both as a narrator (ficto:Narrator) and as a character (ficto:Character), reflecting a homodiegetic intradiegetic narrative configuration, which is represented in the data model via the property ficto:hasNarrativeFunction.

To illustrate how these interrelated components are operationalized within the ontology, we continue with the same example introduced above. The narrative segment narr_seg_expr1.3 is modelled as having the form of speech “direct speech”, specified through the property ficto:hasFormOfSpeech, and is linked to the narrative instance narrator_1.1 via ficto:hasNarrativeInstance. More specifically, the narrative instance corresponds to a homodiegetic narrator (ficto:hasNarrativePosition: “homodiegetic”), that operates on an intra-diegetic level (ficto:hasNarrativeLevel: “intradiegetic”), and adopts an internal point of view (ficto:hasPointOfView: “internal”). In this configuration, it is inferred that the speaker—who addresses the audience through direct discourse—is Dr. Watson himself, a character modelled as an instance of frbroo:F38_Character, and semantically linked to narrator_1.1 via the property ficto:hasNarrativeFunction. This alignment reflects the ontological encoding of a homodiegetic intradiegetic narrator who is simultaneously a character in the fictional world and a narrative instance within the narrative structure.

For the third requirement, regarding the representation of fictional characters, the model reuses the FRBRoo class frbroo:F38_Character. This construct is anticipated to be migrated into the CRMsoc model currently under development (Beretta and Bruseker 2022). A key property associated with this class is R57_is_based_on, which represents fictional characters explicitly derived from real historical or biographical figures. To expand on this, a custom symmetric property called ficto:isInspiredBy has been introduced. This property accounts for the influence between fictional entities, allowing the model to express character-to-character inspiration across different works or traditions. Finally, two additional aspects of the modelling focus on character roles (expressed via dul:hasRole) and character attributes (ficto:Attributes). As discussed in Section 2.1, the formalization of roles is based on the actantial model developed by Greimas (1973), which revisits and refines Propp’s spheres of action (Propp 1966). Specifically, the model identifies three pairs of actants: (1) subject/object, (2) sender/receiver, and (3) helper/opponent.

The proposed model, illustrated in Figure 3, addresses the third requirement by incorporating a detailed set of character attributes derived from the comprehensive analytical framework developed by Marchese (1983). These attributes are formally associated with the fictional character and are semantically linked to both their source of citation (cito:cites) and the specific character instance to which they belong (ficto:hasAttribute), drawing upon Marchese’s identification of five primary anthropological dimensions that characterize fictional entities: (1) Anagraphical Status (ficto:AnagraphicalStatus), encompassing traits such as age, gender, name, physiognomy, socio-economic condition, and professional role; (2) Character and Psychology (ficto:CharacterAndPsychology), which pertains to emotional life, affective tendencies, psychological profile, and self-perception of the character; (3) Axiology (ficto:Axiology), referring to the character’s values and worldviews, including ethical, religious, and political orientations, as well as existential beliefs and attitudes toward others; (4) Praxis (ficto:Praxis), describing actions, behaviors, habits, and communicative patterns such as manner of speaking, idiolect, and gestures; and (5) Modality (ficto:Modality), which addresses both will—such as needs, desires, intentions, and duties—and power, which includes physical means, tools, competences, and specific skills. In addition to these five dimensions, the model introduces a sixth: (6) Symbols (ficto:Symbols). This dimension accounts for the presence of recurring objects or motifs in the character’s fictional experience, along with the symbolic meanings attributed to them by the author.

In addressing the contextual information related to the attributes of characters, two additional properties were introduced. The first property, ficto:linkedTo, is defined as a symmetric property that captures the bidirectional semantic relationship between character-level traits—such as symbolic motifs or praxis-related behaviors—and the conventions of a specific genre. Rather than implying a rigid taxonomy, ficto:linkedTo allows the ontology to express how certain attributes serve as genre markers, reinforcing stylistic patterns and interpretive expectations. This is particularly relevant in genres like detective fiction, where emblematic features (e.g., the magnifying glass or methods of inquiry) are constitutive of both character identity and genre recognition, as demonstrated by the SPARQL queries Q5 and Q6 (see Section 3, Q5 and Q6). The second property, ficto:hasTextualTransmission, is used to encode the transmission of character traits within the narrative tradition. It utilizes a controlled vocabulary that distinguishes between original and apocryphal traits. Attributes identified in the three canonical case studies (see Section 2.1) are classified as original, while those found in the two 20th-century rewritings are considered apocryphal.

To clarify how these ontological mechanisms operate in practice, we may consider another instance drawn from the narrative segment narr_seg_expr2.6, which contains the following passage: “Più un particolare sembra grottesco e stravagante, più vale la pena di esaminarlo attentamente, e proprio ciò che apparentemente complica un caso, se analizzato a fondo e vagliato con metodo scientifico, diventa elemento che finisce per chiarirlo”. This utterance articulates a core epistemological principle of the Holmesian method: the idea that what appears grotesque or excessively complex may become the key to resolving a case. Within the model, this excerpt is classified as an instance of ficto:Axiology, specifically reflecting the epistemological and methodological values that underpin Holmes’s investigative mindset. In the ontology, this axiological trait is semantically linked to the concept labelled “Method” and instantiated as a ficto:Praxis, thereby connecting an internalized value to a specific narrative action. The trait is further associated with the genre of detective fiction through the symmetric property ficto:linkedTo, indicating that such a methodological stance functions as both a character-defining element and a genre-defining convention. Finally, its recurrence in Doyle’s canonical works enables the model to classify it as an “original trait” of Sherlock Holmes via ficto:hasTextualTransmission, aligned with the Getty AAT entry http://vocab.getty.edu/page/aat/300015644 (accessed on 19 August 2025).

Finally, to support the consistency of the semantic framework and facilitate integration across systems, the ontology models each attribute-related entity as an instance of skos:Concept, while the semantic relationships among these concepts are modelled using skos:related. Cross-referencing with external knowledge bases is further implemented through skos:closeMatch for cases of partial equivalence, and owl:sameAs when full semantic identity is established.

As a final step, Figure 4 expands the previous examples by showing how symbolic imagery and interrelated attributes are semantically formalized within the same modelling framework. This use case involves the formal representation of the fictional character in Le loup chantant de Forest Gate. This apocryphal narrative (classified as crm:P2_type and under ficto:apocryphal) centers around Sherlock Holmes, who serves as the main character and is assigned the role of subject (dul:hasRole, ficto:subject) within the underlying narrative structure. The story is narrated by a homodiegetic narrator (ficto:NarrativePosition, ficto:homo-diegetic), who is both part of the story and interacts with other characters on the intra-diegetic level (ficto:NarrativeLevel, ficto:intra-diegetic). This narrator adopts an internal focalization (ficto:Focalization, ficto:foc-internal), presenting the events from a subjective and limited point of view.

This analysis focuses on modelling a significant character trait represented by the metaphor of the “hunting dog”, a recurring symbolic motif throughout various textual representations of Holmes. Originally introduced in Conan Doyle’s canonical corpus and later reappropriated by the OuLiPoPo authors, this metaphor functions as a symbolic attribute (ficto:Symbols) that exemplifies both semantic continuity and intertextual persistence across literary reinterpretations. The metaphorical image of the hunting dog is first developed by Doyle—through the narrative voice of Dr. Watson—on two distinct occasions in A study in Scarlet:

Mi ricordava un bracco di razza ben addestrato che scatta a destra e a manca nel sottobosco (Doyle 2015b, p. 41).

Ma prima che potesse passare del tutto Gregson, Lestrade e Holmes gli balzarono addosso come tanti cani da caccia (Doyle 2015b, p. 77).

This metaphor is then explicitly reactivated in the OuLiPoPo reworking, where it is used to describe Holmes’s investigative method:

Holmes ressemblait maintenant au chien de chasse sur la piste du gibier. Sans plus se soucier de Lestrade ni de ses agents, il s’aplatit sur le tapis, flairant positivement le moindre recoin de la pièce. […] Puis il considéra longuement la surface de marbre qui couvrait celle-ci, allant jusqu’à la mesurer à l’aide du mètre pliant qui ne le quittait jamais.

(OULIPOPO 1975, p. 177)

This passage underscores the metaphorical connection between the detective and the hunting dog, illustrating how Holmes’s investigative behavior is framed through an animalistic lens. It further emphasizes the coexistence of instinct and rationality in his approach: the act of scent-tracking is immediately followed by the meticulous measurement of the scene, suggesting that intuitive perception and deductive reasoning are not opposing forces, but rather interdependent traits that define Holmes’s investigative identity. In the ontological model, this symbolic trait is instantiated as a ficto:Attribute, further specified as ficto:Symbols, and formally linked to the detective genre through the symmetric property of ficto:linkedTo. To represent its literary provenance, the property ficto:hasTextualTransmission is employed to trace the transmission of this trait and classify it as original (see Section 3, Q7), given its presence in the foundational canonical corpus. Finally, the attribute is semantically enriched by modelling it as a skos:Concept, allowing its conceptual alignment with the external knowledge bases such as Wikidata through skos:closeMatch. This layered formalization not only clarifies the internal logic of character construction but also enhances the ontological interoperability of the model, facilitating its application across diverse narrative corpora.

Part of the scenario presented in Figure 4, concerning the description of Le loup chantant de Forest Gate’s diegetic dimension and the symbolic trait with its external reconciliation with Wikidata is formalized in Turtle syntax in Listing 1 as follows:

Listing 1. RDF/Turtle snippet illustrating part of the encoding of Le loup chantant de Forest Gate by OuLiPoPo.

<https://w3id.org/ficto/ExpressionCreation/le_loup_chantant_exprcr>    a lrmoo:F28;    rdfs:label "Le loup chantant de Forest Gate"^^xsd:string;    crm:P2 "http://vocab.getty.edu/page/aat/300404309"^^xsd:anyURI;    crm:P4 "1987"^^xsd:gYear;    crm:P94 <https://w3id.org/ficto/Narrative/narrative_4>. <https://w3id.org/ficto/Narrative/narrative_4>    a ficto:Narrative;    dul:hasPart <https://w3id.org/ficto/NarrativeSegment/narr_seg_expr_4.21>. <https://w3id.org/ficto/NarrativeSegment/narr_seg_expr_4.21>    a ficto:NarrativeSegment;    cito:cites ficto:symb_imag_1. ficto:symb_imag_1    a ficto:Symbols;    rdfs:label "Hunting dog"^^xsd:string;    skos:related <https://w3id.org/ficto/Concept/symbolic_imagery>;    ficto:hasTextualTransmission "http://vocab.getty.edu/page/aat/300015644"^^xsd:anyURI.

3. Results

This section outlines the qualitative and quantitative outcomes of applied case studies, focusing on literary character and the dynamics of intertextual variation through the lens of Semantic Web technologies. During the validation phase of the proposed model, a series of Competency Questions (CQs)—that is, high-level, domain-specific queries—was designed to assess the expressiveness, reasoning capabilities, and inferential potential of the RDF-based knowledge graph built upon the FiCT-O ontology. Developed in parallel with the close reading and semantic modelling of the textual corpus (see Section 2.1), these CQs function as a bridge between narratological analysis and formal semantic representation. Designed to evaluate the model’s capacity to reveal both explicit and implicit relationships among narrative elements, characters, and textual variants, the queries place particular attention to emphasis on how intertextual phenomena manifest across different rewritings of the same fictional entity—in this case, Sherlock Holmes. Framed as interpretative queries in natural language and grounded in the fabula and the plot structure of the selected texts, the CQs are subsequently translated into formal SPARQL queries, enabling direct interrogation of the semantic database. Through this dual articulation—narrative-driven and computational—automated reasoning over structured literary data becomes possible, facilitating knowledge discovery that emerges from the ontological abstraction of narrative elements.

Each of the seven CQs presented below serves as a specific test of the model’s semantic depth and interoperability, while simultaneously serving as a conceptual cross-check: by operationalizing interpretative hypotheses as formal queries, the system demonstrates its capacity to detect patterns, contrasts, and continuities between canonical and apocryphal sources, even when distributed across different linguistic, temporal, and cultural contexts. For the sake of clarity and reproducibility, each CQ is articulated in four components: (1) the query expressed in natural language; (2) a brief contextual framing that explains its rationale and relevance; (3) the corresponding SPARQL translation; (4) a tabular presentation of the retrieved results. A more extensive critical reflection on the epistemological implications of the findings—both in terms of character modelling and in relation to broader literary-theoretical concerns—is developed in the Discussion section (see Section 4), where we explore the interpretive insights enabled by this computational and ontological approach.

• Q1: How many attributes are apocryphal, and to which class do they belong?
  This question investigates the distribution of character attributes that emerge specifically in those apocryphal portions of the corpus that deviate from the canonical tradition. This investigation is conducted through semantic alignment with the Getty AAT concept [http://vocab.getty.edu/page/aat/300404309 (accessed on 19 August 2025)], which is a controlled vocabulary term used to classify non-canonical or derivative works. By classifying each attribute according to its ontological type, the query aims to identify which conceptual domains, such as axiology, psychology, or praxis, are most subject to reinterpretation, expansion, or transformation across rewritings. This approach allows us to assess how later or non-original versions of the text contribute to the evolution of character construction through intertextual variation.

SELECT ?attributetype (COUNT(?attribute) AS ?numOccurrences)

WHERE {

 ?narrseg a ficto:NarrativeSegment.

 ?narrseg prov:value ?value.

 ?narrseg cito:cites ?attribute.

 ?attribute a ?attributetype.

 ?attributetype rdfs:subClassOf ficto:Attribute.

?attribute ficto:hasTextualTransmission "http://vocab.getty.edu/page/aat/300404309"^^xsd:anyURI.

}

GROUP BY ?attributetype

ORDER BY ?attributetype

	attributetype	numOccurrences
1	ficto:Axiology	“6”^^xsd:integer
2	ficto:CharacterAndPsychology	“8”^^xsd:integer
3	ficto:Modality	“1”^^xsd:integer
4	ficto:Praxis	“11”^^xsd:integer
5	ficto:Symbols	“3”^^xsd:integer

2.

Q2: How many attributes are original to which class do they belong?

This question complements Q1 by shifting the focus to the attributes that originate in the canonical core of the corpus—that is, in those narrative segments considered part of the original textual layer. These segments are identified through semantic alignment with the Getty AAT concept [http://vocab.getty.edu/page/aat/300015644 (accessed on 19 August 2025)], which defines the category of “original works” within a controlled vocabulary framework. By classifying these attributes according to their ontological type, the query establishes a baseline semantic profile of the character. This profile serves as a reference model to assess the impact of later reinterpretations. It allows us to measure how much subsequent versions preserve, elaborate on, or subvert the character’s initial narrative configuration.

SELECT ?attributetype (COUNT(?attribute) AS ?numOccurrences)

WHERE {

 ?narrseg a ficto:NarrativeSegment.

 ?narrseg prov:value ?value.

 ?narrseg cito:cites ?attribute.

 ?attribute a ?attributetype.

 ?attributetype rdfs:subClassOf ficto:Attribute.

?attribute ficto:hasTextualTransmission "http://vocab.getty.edu/page/aat/300015644"^^xsd:anyURI.

}

GROUP BY ?attributetype

	attributetype	numOccurrences
1	ficto:AnagraphicalStatus	“22”^^xsd:integer
2	ficto:Axiology	“16”^^xsd:integer
3	ficto:CharacterAndPsychology	“32”^^xsd:integer
4	ficto:Modality	“52”^^xsd:integer
5	ficto:Praxis	“76”^^xsd:integer
6	ficto:Symbols	“18”^^xsd:integer

3.

Q3: Does the psychological dichotomy of apathy and hyperactivity, introduced in A Study in Scarlet, persist across later textual transmissions?

This question focuses on a recurrent psychological contrast that defines Sherlock Holmes’s characterization from the very beginning of the canon—specifically, the tension between apathy and hyperactivity. By examining references to this duality across various narrative segments and expressions, the inquiry investigates the consistency of this trait in different rewritings and adaptations. The goal is to determine whether certain psychological aspects of the character are preserved, reinterpreted, or disrupted over time, thereby providing insight into the ongoing continuity of narrative identity.

SELECT DISTINCT ?exprcr ?narrseg ?value ?attribute

WHERE {

  ?exprcr a lrmoo:F28.

  ?exprcr crm:P94 ?narrative.

  ?narrative a ficto:Narrative.

  ?narrative dul:hasPart ?narrseg.

  ?narrseg a ficto:NarrativeSegment.

  ?narrseg cito:cites ?attribute.

  ?narrseg prov:value ?value.

  ?attribute a ?attributetype.

  ?attribute skos:related ?concept.

$$

  FILTER(?concept IN (

    <https://w3id.org/ficto/Concept/apathy>,

    <https://w3id.org/ficto/Concept/hyperactivity>

  ))

}

	exprcr	narrseg	value	attribute
1	ficto:ExpressionCreation/studio_rosso_exprcr	ficto:NarrativeSegment/narr_seg_expr_1.52	“«S’infilo in fretta il soprabito, muovendosi con una celerità che dimostrava come alla fase di abulia fosse subentrata una crisi di iperattivismo»”	ficto:psych_prof_5
2	ficto:ExpressionCreation/studio_rosso_exprcr	ficto:NarrativeSegment/narr_seg_expr_1.9	“«Non c’era sforzo superiore alle sue energie quando lo prendeva un accesso di attivismo, ma ogni tanto cadeva preda di una reazione contraria e rimaneva sdraiato per giorni di fila sul divano del salotto»”	ficto:psych_prof_5
3	ficto:ExpressionCreation/le_loup_chantant_exprcr	ficto:NarrativeSegment/narr_seg_expr_4.13	“«Et Holmes retomba dans une apathie analogue à celle qui a dû saisir Wellington et toute l’Angleterre lorsque nous eûmes cloué Napoléon au rocher de Saint-Hélèn»”	ficto:psych_prof_5
4	ficto:ExpressionCreation/le_loup_chantant_exprcr	ficto:NarrativeSegment/narr_seg_expr_4.17	“«Je trouvai à Sherlock Holmes des allures de félin tandis qu’il marchait de long en large dans un état de fébrilité qui contrastait avec son abattement de l’heure précédente»”	ficto:psych_prof_5
5	ficto:ExpressionCreation/studio_rosso_exprcr	ficto:NarrativeSegment/narr_seg_expr_1.52	“«S’infilo in fretta il soprabito, muovendosi con una celerità che dimostrava come alla fase di abulia fosse subentrata una crisi di iperattivismo»”	ficto:psych_prof_4
6	ficto:ExpressionCreation/studio_rosso_exprcr	ficto:NarrativeSegment/narr_seg_expr_1.9	“«Non c’era sforzo superiore alle sue energie quando lo prendeva un accesso di attivismo, ma ogni tanto cadeva preda di una reazione contraria e rimaneva sdraiato per giorni di fila sul divano del salotto»”	ficto:psych_prof_4
7	ficto:ExpressionCreation/le_loup_chantant_exprcr	ficto:NarrativeSegment/narr_seg_expr_4.17	“«Je trouvai à Sherlock Holmes des allures de félin tandis qu’il marchait de long en large dans un état de fébrilité qui contrastait avec son abattement de l’heure précédente»”	ficto:psych_prof_4

4.

Q4: Is the idiomatic expression “Elementary, my dear Watson!” present in Doyle’s texts, or is it apocryphal?

This question investigates one of the most iconic expressions attributed to Sherlock Holmes: the idiomatic phrase “Elementary, my dear Watson!”, widely entrenched in the collective imagination. The query aims to determine whether this expression is present in the canonical works of Arthur Conan Doyle or whether it emerged through apocryphal rewritings and adaptations. By tracing its occurrence across both original and non-original narrative segments, the question assesses the extent to which cultural memory and popular reception are grounded in textual evidence or shaped by intertextual reconstruction and reception-driven invention.

SELECT distinct ?exprcr ?narrseg ?attribute ?value ?transmission

WHERE {

 ?exprcr a lrmoo:F28.

 ?exprcr crm:P94 ?narrative.

 ?narrative a dul:Narrative.

 ?narrative crm:P46 ?narrseg.

 ?narrseg a ficto:NarrativeSegment.

 ?narrseg cito:cites ?attribute.

 ?narrseg prov:value ?value.

 ?attribute a ?attributetype.

 ?attribute skos:related <https://w3id.org/ficto/Concept/idiom_expression

 ?attribute ficto:hasTextualTransmission ?transmission.

}

	exprcr	narrseg	attribute	value	transmission
1	ficto:ExpressionCreation/problema_finale_exprcr	ficto:narr_seg_expr_2.4	ficto:expression_1	“«Interessante, per quanto elementare,» concluse, tirando a sedersi nel suo angolo preferito del divano»”	“http://vocab.getty.edu/page/aat/300015644”^^xsd:anyURI
2	ficto:ExpressionCreation/le_loup_chantant_exprcr	ficto:narr_seg_expr_4.18	ficto:expression_2	“«Elémentaire, mon cher Watson!»”	“http://vocab.getty.edu/page/aat/300404309”^^xsd:anyURI
3	ficto:ExpressionCreation/verità_d_exprcr	ficto:narr_seg_expr_5.2	ficto:expression_3	“«- Elementare, mio caro Dupin!—sorride Holmes, seccato in realtà di essere stato prevenuto»”	“http://vocab.getty.edu/page/aat/300404309”^^xsd:anyURI

5.

Q5: How are praxis-related traits distributed across different literary genres?

This question investigates the distribution of praxis-oriented character traits—that is, elements related to action, behavior, gestures, and stylistic embodiment—across distinct literary genres. By correlating praxis attributes with genre classifications, the query seeks to reveal patterns of narrative conduct and expressive stylization that contribute to the construction of genre-specific conventions. The results provide insight into how characters act, behave, or present themselves differently depending on the narrative tradition in which they are inscribed, and how certain gestures or performative codes may serve as genre indicators.

SELECT ?attribute ?genre ?narrseg ?value

WHERE {

  ?narrseg cito:cites ?attribute.

  ?narrseg prov:value ?value.

  ?attribute a ficto:Praxis.

  ?attribute ficto:linkedTo ?genre.

}

	attribute	genre	first_narrseg	first_value
1	ficto:expression_1	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_2.4	“«Interessante, per quanto elementare,» concluse, tirando a sedersi nel suo angolo preferito del divano»”
2	ficto:expression_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_4.18	“«Elémentaire, mon cher Watson!»”
3	ficto:pers_style_7	ficto:LiteraryGenre/fairy_tales	ficto:NarrativeSegment/narr_seg_expr_4.27	“«Holmes ridiculement accoutré d’une pèlerine rouge et d’une petite jupe»”
4	ficto:pers_style_8	ficto:LiteraryGenre/fairy_tales	ficto:NarrativeSegment/narr_seg_expr_4.27	“«Holmes ridiculement accoutré d’une pèlerine rouge et d’une petite jupe»”
5	ficto:expression_3	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_5.2	“«- Elementare, mio caro Dupin!—sorride Holmes, seccato in realtà di essere stato prevenuto»”

6.

Q6: Which symbolic character traits are associated with one or more literary genres?

This question focuses on recurring symbolic traits—such as emblematic objects (e.g., the magnifying glass) and codified gestures—that contribute to the visual and cultural construction of the fictional character. By examining the relationship between these symbolic elements and the literary genres in which they appear, the query explores how such motifs function as genre markers, reinforcing stylistic expectations and interpretive frameworks. The findings highlight how semantic objects come to be associated with specific literary traditions, suggesting that certain physical or material signs can serve as anchors of continuity in the portrayal of the detective figure across canonical and apocryphal texts.

SELECT ?attribute ?genre ?narrseg ?value

WHERE {

?narrseg cito:cites ?attribute.

?narrseg prov:value ?value.

?attribute a ficto:Symbols.

?attribute ficto:linkedTo ?genre.

}

$$

GROUP BY ?attribute ?genre ?narrseg ?value

	attribute	genre	first_narrseg	first_value
1	ficto:recur_p_obj_1	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_1.56	“«Non aveva ancora finito di parlare che tirò fuori di tasca un metro a nastro e una grossa lente d’ingrandimento»”
2	ficto:recur_p_obj_1	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_4.23	“«Puis il considéra longuement la surface de marbre qui couvrait celle-ci, allant jusqu’à la mesurer à l’aide du mètre pliant qui ne le quittait jamais»”
3	ficto:recur_p_obj_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_1.56	“«Non aveva ancora finito di parlare che tirò fuori di tasca un metro a nastro e una grossa lente d’ingrandimento»”
4	ficto:recur_p_obj_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_2.3	“«Poi, con un’espressione di vivo interesse, posò la sigaretta, si avvicinò alla finestra e lo osservò nuovamente con una lente di ingrandimento»”
5	ficto:recur_p_obj_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_4.22	“«sa loupe»“
6	ficto:recur_p_obj_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_5.1	“«Stanno poi sorbendo i loro cappuccini quando un uomo alto, ben riconoscibile dal mantello a larghe pieghe e dal singolare berretto, viene a esaminare le loro targhette con una grossa lente—Permettete? Holmes,—dice presentando anche il suo compagno, sul risvolto del quale si legge: «Dr. Watson»”
7	ficto:recur_p_obj_2	ficto:LiteraryGenre/detective_fiction	ficto:NarrativeSegment/narr_seg_expr_5.3	“«[…] per dare la parola a Holmes, che ha alzato la sua lente.—C’è—dice l’esperto di cani fluorescenti,—un puntino che vorrei mettere subito su una certa i […]»”

7.

Q7: What types of symbolic traits recur across the corpus, and how frequently do they appear?

This question explores the typology and frequency of symbolic traits associated with fictional characters across the entire corpus. These traits include both recurring physical objects (such as the magnifying glass or tape measure) and symbolic metaphors or expressions (such as “hunting dog” or “devil of a man”), each of which contributes to the semantic construction of character identity and genre coherence. By quantifying the number of occurrences of each symbol and grouping them according to their ontological label, the query highlights which motifs are most persistent across canonical and apocryphal texts. This allows us to assess how symbolic recurrence functions as a mechanism of narrative stabilization, fostering continuity of interpretation and recognizability within the evolving fictional universe.

SELECT ?attribute ?label (COUNT(?attribute) AS ?numOccurrences)

WHERE {

 ?narrseg a ficto:NarrativeSegment.

 ?narrseg cito:cites ?attribute.

 ?attribute a ficto:Symbols.

 ?attribute rdfs:label ?label.

}

GROUP BY ?attribute ?label

	attribute	genre	first_narrseg	first_value
1	ficto:recur_p_obj_1	“Tape measure”	“2”^^xsd:integer	ficto:recur_p_obj_1
2	ficto:recur_p_obj_2	“Magnifying glass”	“5”^^xsd:integer	ficto:recur_p_obj_2
3	ficto:symb_imag_1	“Hunting dog”	“3”^^xsd:integer	ficto:symb_imag_1
4	ficto:symb_imag_2	“Bloodhound”	“1”^^xsd:integer	ficto:symb_imag_2
5	ficto:symb_imag_3	“Lark”	“1”^^xsd:integer	ficto:symb_imag_3
6	ficto:symb_imag_4	“General”	“1”^^xsd:integer	ficto:symb_imag_4
7	ficto:recur_p_obj_3	“Pipe du matin (morning pipe)”	“3”^^xsd:integer	ficto:recur_p_obj_3
8	ficto:recur_p_obj_4	“Silver cigarette case”	“1”^^xsd:integer	ficto:recur_p_obj_4
9	ficto:symb_imag_5	“Diable d’homme (devil of a man)”	“1”^^xsd:integer	ficto:symb_imag_5
10	ficto:symb_imag_6	“Feline-like manner”	“1”^^xsd:integer	ficto:symb_imag_6
11	ficto:recur_p_obj_5	“Persian babouche (Persian slippers)”	“1”^^xsd:integer	ficto:recur_p_obj_5
12	ficto:symb_imag_7	“Expert in fluorescent dogs”	“1”^^xsd:integer	ficto:symb_imag_7

4. Discussion

Grounded in a rigorous close reading of the selected corpus (see Section 4.1), our investigation adopted a semantic-quantitative methodology, using SPARQL queries over the RDF-based knowledge graph to validate the ontological model and explore the evolution of the fictional character of Sherlock Holmes. This section reflects critically on the epistemological implications of the results presented above, framing them in light of the overarching RQ: How has the representation of the detective figure evolved in detective fiction from the 19th to the 20th century?

Focusing on a corpus centered on Sherlock Holmes and his apocryphal rewritings (see Section 4.1), the data retrieved through our CQs suggest a general tendency in 20th-century intertextual adaptations to maintain fidelity to the core fictional universe originally constructed by Arthur Conan Doyle. As illustrated by the results of CQ1 and CQ2, the number of apocryphal attributes introduced in later texts is significantly lower than the canonical attributes emerging from the original trilogy. Notably, apocryphal traits do not intervene on biographical aspects of the character (e.g., age, gender, and socio-professional identity), but rather target dimensions of subjectivity and narrative behavior, especially those related to psychology, praxis, and modality.

This asymmetry is particularly meaningful: it reveals a semiotic hierarchy among the different character traits, distinguishing those considered immutable (co-constitutive of Holmes’s fictional identity) from those more readily subject to variation and experimentation. Canonical features, such as Holmes’s epistemic posture, axiological system, and core biography, appear to function as narrative constants that maintain diegetic coherence. By contrast, stylistic idiosyncrasies, habits, and gestural expressions are more susceptible to transformation and offer a fertile space for genre play and metafictional operations. This semiotic asymmetry not only delineates a hierarchy of character traits but also suggests a potential critical function in the apocryphal rewritings, where peripheral traits are strategically amplified to comment upon or subvert the canonical model.

A paradigmatic example of this dynamic is offered by CQ4, which investigates the origins of the idiomatic phrase “Elementary, Watson!”. Despite its iconic status within the Sherlockian imaginary, the phrase is absent in Doyle’s corpus. Instead, our analysis traces its genesis to a formulation in A Study in Scarlet, later reworked in 20th-century pastiches. This case exemplifies how collective memory, rather than strictly adhering to textual evidence, operates through iterative processes of citation, recontextualization, and cultural crystallization. Further evidence of this interpretive stratification is offered by CQ3, which examines the psychological polarity between apathy and hyperactivity introduced in Doyle’s early works. This trait reappears explicitly in later adaptations—such as Le Loup chantant de Forest Gate—highlighting how certain narrative motifs become structural invariants of the character, even when filtered through experimental literary forms like pastiche and parody.

Another layer of interpretive divergence emerges from the analysis of narrative voice and focalization strategies. Le Loup chantant de Forrest Gate adheres more closely to Doyle’s model: the narrator is homodiegetic, shifting between internal and zero focalization across intra- and extradiegetic levels. This configuration mirrors the function of Dr. Watson in the canonical corpus, acting as an interpretive proxy who modulates access to the diegetic world and maintains epistemic asymmetry. In contrast, La verità sul caso D. exhibits a more radical metanarrative experimentation: when Holmes appears, the narrator adopts a heterodiegetic stance—external to the story and aligned with zero focalization. This formal reconfiguration reinforces the pastiche’s reflexive dismantling of narrative coherence and genre expectations, foregrounding the mechanisms of narration itself as an object of parody and critique.

The interdependence between character modelling and genre conventions also emerges as a central insight from our model. In particular, CQ5 and CQ6 demonstrate that some traits—especially those involving symbols and narrative praxis—act as genre markers, reinforcing interpretive expectations. Using the property ficto:linkedTo, we could isolate genre-specific configurations of Holmesian behaviors, such as the ritual use of the magnifying glass, or certain verbal stylizations that function as semiotic anchors within the genre of detective fiction. This observation aligns with Eco’s notion of the library of possible worlds (Eco 2002), wherein each fictional world is semiotically inscribed within a larger intertextual network. The results of CQ5 illustrate this point vividly, showing how symbolic attributes of Holmes are reframed within mythopoetic templates drawn from fairy tale tradition, such as the explicit echo of Little Red Riding Hood.

Ultimately, the discussion of these findings highlights the expressive potential of semantic technologies to formalize literary interpretation. By leveraging RDF modelling and SPARQL querying, our approach enables a granular, data-driven understanding of narrative structures, character constellations, and genre dynamics. This not only validates the epistemic value of ontological modelling in literary studies but also opens new avenues for scalable, interoperable, and hermeneutically grounded analysis of fictional entities. Thanks to the alignment with external vocabularies such as the Getty AAT, the dataset ensures semantic interoperability and contributes to the broader ecosystem of Linked Open Data in the humanities.

5. Conclusions

This study aims to explore the epistemic potential of formal models of knowledge organization when tested within a narratological approach to the interpretation of fictional texts. The application of Semantic Web methodologies makes it possible to highlight the inherent arbitrariness of the creative act—the fictional world as a “possible state of affairs” (Eco 1979)—and to provide scholars with a transparent and expressive vocabulary for identifying the levels at which this arbitrariness operates. Following Eco’s formulation, the fictional world becomes a model of propositions that denote it, while the ontological model traces these propositions, aiding in the comprehension of the content they express. In this regard, even the “true problem of narratology” (Marchese 1983), namely the status of the character, is addressed by providing an expressive taxonomy that qualifies it “as an individual, along with its defining properties” (Eco 1979). Building on the reflections of Chatman (1978) and Segre, this approach bridges the gap between the character as a role and as a logical function, conceptualizing it as a “bundle of attitudes and character traits” (Segre 1974), which substantiates its motivations and sphere of action.

At the same time, we remain aware of the interpretative variability intrinsic to humanistic inquiry. The use of a formal model grounded in robust theoretical foundations—specifically, the analytical framework of narratology—does not eliminate this variability but rather provides a shared conceptual architecture through which it can be meaningfully articulated. The aim of this study is not to constrain interpretation, but to test whether a semantically structured model can support and withstand diverse critical perspectives. The outcomes are promising: by anchoring interpretation in a formal model and linking it to reusable ontological structures, we create the conditions for a critical dialogue between modelling and interpretation. As interpretative disagreements emerge—e.g., the psychological reading of a trait or the classification of a narrative perspective—our model is designed to accommodate alternative viewpoints. This is made possible by the Semantic Web’s inherent interoperability and modularity: while maintaining a consistent logical backbone, FiCT-O can be extended to incorporate additional perspectives and contextual layers.

A particularly fruitful trajectory involves the integration of ontologies such as HiCO (Historical Context Ontology15), which is specifically designed to model interpretative acts and provenance. This enables the tracing of divergent interpretive claims within a coherent and semantically expressive system, supporting a hierarchy of assertions that preserves consistency while capturing hermeneutic richness. Additionally, starting from the formalizing statements already documented with the property prov:value, the use of provenance ontologies further enhances the trustworthiness and transparency of the model, especially when dealing with interpretative pluralism.

From a methodological standpoint, the modular approach adopted in this study—combining close reading with the construction of expressive semantics—opens several future directions. As emphasized throughout this article, the semantic web framework supports the “scalable integration” of close and distant reading techniques, enabling comparative analysis across multiple levels of granularity. Figure 5 illustrates this potential: by visualizing the relationship between narrative propositions (ficto:NarrativeSegment) and character-related property classes, we obtain a synthesis of quantitative and qualitative insights that illuminates how the text constructs character identity through interaction with the surrounding environment (especially through the categories of Praxis and Modality). This reinforces Eco’s assertion that a fictional world, as “a possible state of affairs expressed through propositions”, also implies “a possible course of events” (Eco 1979).

The reflections on the potential of complementary approaches to the methodology identified in this study simultaneously define the further trajectories and next steps of the research. First, from a structural perspective, we intend to expand the model to encompass the full network of actantial relationships involving the protagonist-subject and other characters—specifically the narrative roles of object, receiver, helper, and opponent (Greimas 1973). These narrative roles must be qualified through semantically rich and scalable properties, just as we did for the detective figure in the current study.

Second, in terms of scope, two parallel objectives arise: validating the model’s consistency and enhancing its reusability. On the one hand, future work will extend the corpus to include the original English texts by Doyle, thereby assessing how the phenomenon of translation affects the model’s structural robustness. On the other hand, the model will be applied to characters from broader literary traditions of the 19th and 20th centuries, expanding the descriptive reach of ontology. This application will take place across a multilingual and multicultural landscape, exploring the model’s performance across literary systems. Within this broader scope, a particularly promising direction is the application of FiCT-O to the semantic analysis of literary genre traditions, using multi-level modelling to compare character construction, narrative perspective, and actantial roles across different texts.

By leveraging the property ficto:hasTextualTransmission, and through the structured modelling of character traits, narrative instances, and symbolic functions, the ontology supports the comparative analysis of transtextuality typologies, as theorized by Genette ([1982] 1989). This allows us to examine how traits and functions migrate, persist, or mutate across time, language, and genre. As previously discussed (Section 2.2), the proposed model integrates high-level ontologies from the cultural heritage sector. This integration facilitates the contextualization of texts and narrative components within broader knowledge infrastructures. Additionally, it enhances the ability to detect patterns of narrative construction and to identify salient traits that define character identity across a corpus. In doing so, FiCT-O offers a reusable, interoperable framework for semantic modelling in literary studies, one that combines interpretive flexibility with formal precision, and that may prove valuable across diverse narrative domains and critical traditions.