Sensorium Conversational Corpus for Retrieval-Augmented Generation

# Sensorium Conversational Corpus for Retrieval-Augmented Generation

## Overview

The **Sensorium Conversational Corpus** is a curated multimodal text dataset compiled for use in the *Sensorium ARC* project — an interactive AI agent system for oceanic data exploration and eco-poetic climate communication.  

This corpus contains PDF documents drawn from eco-art archives, scientific reports, project frameworks, and transcribed poetic dialogues. It was used as the **primary Retrieval-Augmented Generation (RAG) database** in the experiments and deployments described in *Sensorium ARC: AI Agent System for Oceanic Data Exploration and Interactive Eco-Art*.

The documents represent:

- **Eco-artistic philosophy** from Newton and Helen Mayer Harrison's works

- **Project documentation** and conceptual frameworks

- **Scientific and environmental data** (including NASA datasets)

- **Transcribed poetic dialogues** with the “Ocean” persona

---

## File List

- `CFM.pdf` – Center for the Study of the Force Majeure overview

- `CRB Community Resilience Web Project Framework.pdf` – Community Resilience Web project framework

- `Force Majeure Statement.pdf` – Statement document on the Force Majeure concept

- `ForceMajeureBook.pdf` – Major Harrison eco-art project text

- `FromThere.pdf` – Autobiographical retrospective

- `Harrison_Sensorium_InEcopoesisJournal.pdf` – Published article on Sensorium

- `HarrisonsCaliforniaWork.pdf` – Regional eco-art works

- `NASA Data.pdf` – NASA dataset documentation

- `PenEuroCatE.pdf` – Peninsula Europe project materials

- `README.md` – Dataset documentation

- `Sensorium.pdf` – Project conceptual overview

- `The Force Majeure Manifesto Tibet Peninsula Europe the Sierra Nevada and Center statement.pdf` – Comprehensive manifesto document

- `[transcript] Apologia Mediterraneo.pdf` – Transcript of poetic dialogue

- `[transcript] Sensorium for the World Ocean.pdf` – Transcript of poetic dialogue

- `websources.pdf` – Collected web sources

---

---

## Usage

This dataset was used to populate a local **vector database** via sentence-level chunking and embedding with the `all-MiniLM-L12-v2` model, enabling retrieval for a domain-specific conversational AI system.

---

## License

**CC BY-NC-ND 4.0** — This dataset may be shared for **non-commercial purposes only**, with attribution, and without modification.  

**Additional requirement:** Any use of this dataset (including academic research) must be reported to the author via the contact email provided in the Zenodo record.

---

## Citation

If you use this dataset, please cite:

> Lee, Myungin. *Sensorium Conversational Corpus (PDF) for RAG*. Zenodo. DOI: [to be assigned]

BibTeX (update after DOI assigned):

```bibtex

@dataset{lee_sensorium_corpus_2025,

  author       = {Myungin Lee},

  title        = {Sensorium Conversational Corpus (PDF) for RAG},

  year         = {2025},

  publisher    = {Zenodo},

  doi          = {10.5281/zenodo.16777214},

  url          = {https://doi.org/10.5281/zenodo.16777214}

}