This repository contains the dataset accompanying the SoundAI paper (2026-01). It includes both experimental datasets and extensive field data collected in Beijing. The field data covers a wide range of urban metrics, including air pollution, meteorological observations, subway mobility, and anti-fraud intervention dialogue results spanning multiple years.

Repository Structure

The repository is organized into three main directories:

experiment_1/: Contains data and results related to the first experiment.
experiment_2/: Contains data and results related to the second experiment.
field_data/: Contains the primary empirical datasets used for real-world analysis.

Field Data Inventory

The field_data directory contains the following files, provided in multiple formats (CSV, Stata .dta, and JSON) for ease of use across different statistical and machine learning frameworks.

Air Pollution & Meteorology

File Name	Format	Time Range	File Size
`Beijing_Air_Pollution_Data_201405_202306`	`.csv`, `.dta`	May 2014 – Jun 2023	~5.6 MB / 3.7 MB
`Beijing_Meteorological_Observation_Data_199001_202412`	`.csv`, `.dta`	Jan 1990 – Dec 2024	~16.7 MB / 13.7 MB

Anti-Fraud Interventions

File Name	Format	Time Range	File Size
`Beijing_Anti_Fraud_Dialogue_Result_202201_202302`	`.csv`, `.dta`	Jan 2022 – Feb 2023	~175.9 MB / 102.6 MB
`Beijing_Anti_Fraud_Intervention_Data_202109_202302`	`.csv`, `.dta`	Sep 2021 – Feb 2023	~358.8 MB / 451.8 MB

Urban Mobility

File Name	Format	Time Range	File Size
`Beijing_Subway_Mobility_201707_202512`	`.json`	Jul 2017 – Dec 2025	~319 KB

Usage Information

Supported Formats:

CSV: Standard comma-separated values, accessible via Python (pandas), R, or Excel.
DTA: Stata dataset format, optimized for econometric and statistical software.
JSON: Structured data format, used here for subway mobility records.

License: This dataset is released under the openrail++ license. Please ensure your use cases comply with the ethical usage guidelines outlined in the OpenRAIL framework.

Downloads last month: 4