> For the complete documentation index, see [llms.txt](https://gang-gems.gitbook.io/qtv/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://gang-gems.gitbook.io/qtv/introduction/knowledge-representation-and-extraction.md).

# Knowledge Representation & Extraction

This project combines two core methodological pillars from Semantic Web research:

1. **Knowledge Representation**\
   Building a formal ontology that models how LGBTQ+ characters are portrayed in teenage dramas.
2. **Knowledge Extraction**\
   Gathering narrative and empirical data from real-world sources to populate and refine the ontology.

Together, these steps allow us to move from qualitative media narratives to structured, computable knowledge.

***

### 🧩 From Stories to Structure

| Phase                  | Activity                                                    | Method                         | Output                                      |
| ---------------------- | ----------------------------------------------------------- | ------------------------------ | ------------------------------------------- |
| **Domain Study**       | Literature review (GLAAD, media theory)                     | Manual analysis                | High-level concepts, terminology            |
| **Narrative Mining**   | Character scenarios and one user story (e.g., Jules Vaughn) | Prompting LLMs + story mapping | Competency questions, portrayal traits      |
| **Dataset Collection** | Compilation of 29 LGBTQ+ characte                           | Manual annotation              | CSV dataset with 29+ attributes             |
| **Extraction**         | From dataset to ontology                                    | Attribute mapping              | Individuals, subclasses, refined properties |

***

### 🏗️ Knowledge Representation: Ontology Design

We followed a two-phase modeling approach:

**Hybrid modeling – Conceptual Modeling (Level I)**\
Our first level of modeling combines top-down and bottom-up reasoning.\
It was based on existing theory and ontologies,  but also informed by user stories and narrative examples.\
→ See 📐 [*Conceptual Modeling – Level I*](/qtv/conceptual-modeling-level-i.md)

**Bottom-up enrichment – Dataset Integration (Level II)**\
The second phase involved refining and expanding the ontology using real data from our annotated character dataset.\
→ See 📊 [*Dataset Integration – Level II*](/qtv/dataset-integration-and-bottom-up-ontology-level-ii.md)

***

### 💡 Why This Matters

This dual approach ensures that:

* The ontology is grounded in both theory and practice
* The model reflects real-world portrayals, not just ideal categories
* The structure supports meaningful research questions about queer visibility, narrative bias, and cultural framing

We don’t just describe the media — we model it as a system of **cultural knowledge**, ready for analysis, critique, and transformation.

***