Speakers
Description
This paper presents a computational narrative dictionary of youth language in contexts of high socio-economic vulnerability, where “narrative” is understood, following Bruner’s distinction between paradigmatic and narrative modes of meaning-making (Bruner, 1986), as meaning grounded in experience, examples, values, and self-positioning rather than only in categorisation. The resource is designed for social and community workers, educators, and psychologists who work with young people in fragile settings and need to interpret not only what words denote, but also the experiences, values, and expectations that speakers attach to them. Its aim is therefore both descriptive and practical: to document situated meanings and make them usable in social and educational intervention. The dictionary does not replace standard lexicographical description; rather, it complements it with information that is usually excluded from dictionary entries but is central to intervention work: emotions, examples, social roles, expectations, and forms of self-positioning.
The dictionary is being developed within the project “Futuri (im)possibili: diagnosing the present and imagining the future” with the youth of Caivano , carried out in the metropolitan area of Naples, Italy. The project adopts an Action-Research approach (Elliot, 1993; Lewin 1946), combining empirical investigation with community-based social intervention. In line with user-oriented lexicography, the dictionary addresses concrete needs: it supports professionals in recognising semantic shifts and resemantisations, and in designing interventions that can help young people imagine alternative and more positive futures.
Building on Bruner’s distinction, the resource focuses on socially sensitive Italian terms connected with power, deviance, gender, violence, and identity, such as ‘aborto’, ‘criminale’, ‘femmina’, ‘infame’, and ‘soldi’. While standard dictionaries usually formalise meanings through paradigmatic definitions based on categorisation, synonymy, and semantic inclusion, the proposed dictionary also records associative networks, lived experience, and identity-related narratives. For instance, the word ‘femmina’ may activate biological or dictionary-like meanings, while also eliciting narratives of strength and care, but also inferiority, sexualisation, and violence. Meaning is thus treated as a cognitive and cultural construct shaped by shared representations, stereotypes, and lived experience.
The empirical basis consists of 58 semi-structured interviews (Adams, 2015), corresponding to 61 recordings and over 30,000 tokens of transcribed speech, collected mainly in street contexts in Caivano by outreach volunteers. Participants were aged 12 to 30, with a smaller group over 30, and were mostly residents of the area. Interviews were audio/video recorded, digitally transcribed, anonymised to ensure privacy protection, and organised in a structured corpus with metadata and unique identifiers. Automated transcription was combined with manual revision to guarantee accuracy. At the time of writing, the dictionary contains 16 lexical entries with both paradigmatic and narrative meanings.
The dictionary has been provided with a computational representation following OntoLex-Lemon , the de facto standard for lexical resources in the Semantic Web. This choice supports FAIR principles (Wilkinson et al., 2016) and facilitates interoperability with resources in the Linguistic Linked Data cloud. Each entry is modelled as a lexicog:Entry linked to an ontolex:LexicalEntry, while meanings are represented as ontolex:LexicalSense. Paradigmatic and narrative senses are encoded as SKOS concepts in a dedicated skos:ConceptScheme and assigned through dc:type. In the case of ‘femmina’ (Figure 1*), the paradigmatic sense is linked to the concept WOMAN, whereas narrative senses are connected to concepts such as STRENGTH and POWER. Evaluative polarity is modelled through the MARL Opinion Ontology by reifying the ontolex:isLexicalizedSenseOf relation.
The model is evaluated through competency questions that reflect expected use cases. These include, for example, retrieving concepts associated with narrative senses for a given entry, listing concepts connected with negative or positive polarity, and identifying all narrative senses related to a specific concept. Rather than treating polarity as a simple label attached to an entry, the model relates it to specific senses and concepts, since the same word may carry different, even conflicting, evaluations depending on the speaker’s narrative frame and social experience.
Future work will connect senses to textual occurrences in the corpus through OntoLex-FrAC and enrich interview metadata using standard vocabularies such as Dublin Core, FOAF, and PROV-O. Integrating lexical, conceptual, textual, and metadata layers will enable cross-layer queries over the dataset, for example to examine which narrative senses and associations emerge in relation to speakers’ age and gender. This can provide evidence-based insights for professionals working with youth in socially fragile contexts.
* see Book of Abstracts