Upload embeddings_with_metadata.npz

# NUFORC UFO Sightings Dataset

## Dataset Summary

This dataset contains a collection of UFO sighting reports from the **National UFO Reporting Center (NUFORC)**, consisting of narratives, embeddings, and metadata. Each record represents an individual UFO sighting and includes the following:

- **Sighting ID**: A unique identifier for each sighting.
- **Narrative**: A cleaned textual description of the sighting.
- **Embeddings**: Precomputed sentence-transformer-based embeddings of the narratives for machine learning tasks.

This dataset is ideal for applications such as natural language processing (NLP), clustering, similarity analysis, and the study of UFO narratives.

---

## Dataset Structure

The dataset is stored as a compressed `.npz` file with the following components:

- **`embeddings`**: A NumPy array of embeddings generated using the `BAAI/bge-small-en-v1.5` model.
- **`sighting_ids`**: A NumPy array of unique sighting IDs corresponding to the embeddings.
- **`narratives`**: A list of cleaned narrative texts corresponding to each sighting ID.

---

## Dataset Creation

- **Source**: The narratives were collected from NUFORC's publicly available UFO reports.
- **Preprocessing**:
- Text narratives were cleaned to remove incomplete or irrelevant entries.
- Embeddings were computed using the `BAAI/bge-small-en-v1.5` sentence transformer model.
- **Limitations**: No additional metadata (e.g., timestamps or geolocations) is included.

---

## Dataset Usage

- **Primary Applications**:
- Clustering and similarity analysis of UFO narratives.
- Anomaly detection and unsupervised learning.
- Training or fine-tuning NLP models on UFO-related data.
- **Example Use Case**:
```python
import numpy as np
npz_data = np.load('embeddings_with_metadata.npz')
embeddings = npz_data['embeddings']
sighting_ids = npz_data['sighting_ids']
narratives = npz_data['narratives']

# Example: Searching for the closest narrative
from sklearn.metrics.pairwise import cosine_similarity
query_embedding = embeddings[0]
similarities = cosine_similarity([query_embedding], embeddings)
most_similar_idx = similarities.argmax()
print(f"Most similar narrative: {narratives[most_similar_idx]}")

embeddings_with_metadata.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebc812a797d9926851148031d7d4480edd811648736550418f1a36ca90617ea2
+size 281866186