Hollywood: Towards a Large Movie Dataset for Database Benchmarking

2026-07-21

AI Summary

Existing IMDb datasets suffer from fixed scale and homogeneous distribution, limiting their utility in evaluating the generalization capabilities of cardinality estimators. This work proposes the first framework that integrates semantic knowledge generated by large language models (LLMs) with a temporal-graph-driven approach to synthesize relational data, enabling the construction of controllable-scale, reproducible, and semantically rich synthetic datasets while preserving compatibility with the original IMDb schema. Using this framework, we generate Hollywood-200K—a dataset comprising 200,000 movies, 19.7 million table rows, and 213 non-zero JOB queries—which achieves cardinality estimation accuracy on two open-source systems that matches or exceeds that of the original IMDb, substantially enhancing the scalability and practicality of benchmarking for query optimization.

Abstract

The IMDb real-world dataset of the JOB benchmark has been extensively used in the last decade as part of the research line on cardinality estimation, given its ability to stress test both traditional and learned estimators. However, unlike the synthetic TPC family, it does not come with a scale factor, being a simple dump. We introduce Hollywood, a synthetic IMDb-compatible benchmark generator that combines LLM-generated semantic dictionaries with deterministic temporal-graph-based relational data generation. We analyze a preliminary Hollywood-200K, which contains 200,000 primary movies, generated series and episode title rows, 19.7M IMDb-style rows, and 213 nonzero JOB-Light, JOB, and JOB-Complex queries. Experiments with two open systems demonstrate that Hollywood induces cardinality estimation errors comparable to or exceeding those observed on the original IMDb dataset. The release includes generation settings and prompt/LLM-output provenance together with adapted SQL and labels, enabling tests of whether cardinality estimators generalize beyond a fixed movie snapshot and distribution.

Problem

• cardinality estimation
• database benchmarking
• IMDb dataset
• scalability
• synthetic data generation

Innovation

• synthetic data generation
• LLM-generated semantic dictionaries
• temporal-graph-based relational data
• cardinality estimation benchmarking
• IMDb-compatible dataset