---
license: mit
task_categories:
- text-classification
language:
- en
tags:
- sentiment-analysis
- multi-label
- quotes
- llm-evaluation
- fine-tuning
size_categories:
- 10K<n<100K
---

# 75k+ Multi-Label Quotes Dataset for Sentiment & Categorization

This dataset contains **75,966 records** of highly curated quotes across multiple labels, engineered explicitly for training text classification layers, instruction fine-tuning evaluation, and benchmarking high-concurrency sentiment analysis workflows.

## 📊 Dataset Schema & Distributions

The dataset consists of 5 distinct features:

*   **`QUOTE`** (`string`): The textual content of the quote.
*   **`AUTHOR`** (`string`): The individual credited with the quote.
*   **`GENRE`** (`string`): Sub-genre contextual tags.
*   **`category`** (`string`): Primary thematic multi-label mapping.
*   **`sentiment`** (`string`): Ground-truth binary sentiment mapping.

### Category Distribution Breakdown:
*   **motivational**: 27,759 rows
*   **sarcastic**: 25,760 rows
*   **funny**: 15,557 rows
*   **love quote**: 6,890 rows

### Sentiment Distribution Breakdown:
*   **positive**: 50,206 rows
*   **negative**: 25,760 rows

---

## 🛠️ Intended Use Cases

1. **Multi-Label Sequence Classification:** Ideal for training encoder networks (like RoBERTa or DistilBERT) to classify incoming texts into nested thematic genres and sentiment targets simultaneously.
2. **Instruction-Tuning Alignment (PEFT):** Formatted for direct structural mapping when adapting open-weight LLMs (like LLaMA 3 8B) to generate highly contextual responses with intentional emotional undertones (e.g., prompting for a sarcastic or motivational persona).
3. **High-Concurrency Benchmarking:** Perfect dataset for batch ingestion into a streaming API or populating vector embedding stores (like Pgvector or Redis) to evaluate retrieval speeds over 75k+ concurrent row reads.

## 🚀 Ingestion Quick Start

```python
from datasets import load_dataset

# Pull the production-ready split from the Hub
dataset = load_dataset("rama9agg/quotes-sentiment-dataset", split="train")

# Convert sample directly into a pandas frame
df = dataset.select(range(5)).to_pandas()
print(df)