Product database for running recommendation experiments with the Personas framework.
This dataset contains product information and categorization data used for conversational recommendation experiments.
Statistics:
products.parquet - Product information (id, title, price, category, raw metadata)categories.parquet - Category names and IDsproduct_category.parquet - Many-to-many links between products and categoriespersona_scores.parquet - Cached LLM scores for persona-product pairsmetadata.json - Export metadata and statistics| Column | Type | Description |
|---|---|---|
| id | int | Internal product ID |
| external_id | str | Amazon ASIN |
| title | str | Product name |
| main_category | str | Amazon's main category |
| store | str | Store name |
| price | float | Price in USD |
| raw | str | Full product metadata (JSON string) |
| Column | Type | Description |
|---|---|---|
| id | int | Category ID |
| name | str | Category name |
| Column | Type | Description |
|---|---|---|
| product_id | int | References products.id |
| category_id | int | References categories.id |
| Column | Type | Description |
|---|---|---|
| persona_index | int | Persona ID (from Tianyi-Lab/Personas) |
| category_id | int | References categories.id |
| product_id | int | References products.id |
| score | float | Persona's score for product (0-100) |
| reason | str | Explanation for score |
| model | str | LLM model used for scoring |
| created_at | int | Unix timestamp |
from datasets import load_dataset
# Load as HuggingFace dataset
dataset = load_dataset("gilberty005/personas-product-database")
# Access individual tables
products = dataset['products'].to_pandas()
categories = dataset['categories'].to_pandas()
# The experiment code includes auto-download and rebuild functionality
# Just run your experiment and the database will be created automatically!
from pipeline.core.simulate_interaction import ensure_database
ensure_database() # Downloads Parquet files and builds SQLite DB
MIT License
Product database for running recommendation experiments with the Personas framework.
This dataset contains product information and categorization data used for conversational recommendation experiments.
Statistics:
products.parquet - Product information (id, title, price, category, raw metadata)categories.parquet - Category names and IDsproduct_category.parquet - Many-to-many links between products and categoriespersona_scores.parquet - Cached LLM scores for persona-product pairsmetadata.json - Export metadata and statistics| Column | Type | Description |
|---|---|---|
| id | int | Internal product ID |
| external_id | str | Amazon ASIN |
| title | str | Product name |
| main_category | str | Amazon's main category |
| store | str | Store name |
| price | float | Price in USD |
| raw | str | Full product metadata (JSON string) |
| Column | Type | Description |
|---|---|---|
| id | int | Category ID |
| name | str | Category name |
| Column | Type | Description |
|---|---|---|
| product_id | int | References products.id |
| category_id | int | References categories.id |
| Column | Type | Description |
|---|---|---|
| persona_index | int | Persona ID (from Tianyi-Lab/Personas) |
| category_id | int | References categories.id |
| product_id | int | References products.id |
| score | float | Persona's score for product (0-100) |
| reason | str | Explanation for score |
| model | str | LLM model used for scoring |
| created_at | int | Unix timestamp |
from datasets import load_dataset
# Load as HuggingFace dataset
dataset = load_dataset("gilberty005/personas-product-database")
# Access individual tables
products = dataset['products'].to_pandas()
categories = dataset['categories'].to_pandas()
# The experiment code includes auto-download and rebuild functionality
# Just run your experiment and the database will be created automatically!
from pipeline.core.simulate_interaction import ensure_database
ensure_database() # Downloads Parquet files and builds SQLite DB
MIT License