Workshop Schedule
| Time | Topic | Author |
|---|---|---|
| 09:00-09:30 | The Story of Dataprep.ai | Jiannan Wang |
| 09:30-09:45 | Addressing Data Management Challenges for Interoperable Data Science | Ilin Tolovski, Tilmann Rabl |
| 09:45-10:00 | Missing Value Imputation via Pre-trained Language Models with Trainable Prompt and Retrieval Augmentation | Xiang Huang, Shuang Hao |
| 10:00-10:10 | Coffee Break | |
| 10:10-10:40 | Accelerating Storage with Polar-Stack on Baidu AI Cloud | Li Shuo |
| 10:40-10:55 | LLM-assisted Labeling Function Generation for Semantic Type Detection | Chenjie Li, Dan Zhang, Jin Wang |
| 10:55-11:10 | Approximate Functional Dependencies Discovery Using Markov Blanket | Jinqi Liu, Anzhen Zhang, Jiajia Li, Na Guo, Jing Zhang |
Abstract
The advent of artificial intelligence (AI), particularly through deep learning (DL) and large language models (LLMs), has marked a significant milestone in technological advancement, attributing to its unparalleled accuracy and generalization abilities. The rapid evolution of AI model structures to achieve superior performance underscores the dynamic progression and potential of AI technologies. However, the cornerstone of any AI's success lies not just in its algorithmic prowess but in the quality of data it is trained on. High-quality, accurate, consistent, and representative data sets are imperative for enhancing AI models' learning efficacy, thereby optimizing their generalization capabilities and reducing computational demands.
Beyond just leveraging quality data, AI technology itself plays a pivotal role in enhancing data quality through its powerful tools for data management. From cleaning, labeling, and validation to sophisticated feature engineering, AI ensures data accuracy, integrity, consistency, and reliability. This creates a symbiotic relationship between AI technology and high-quality data, highlighting their mutual dependence and the complementary nature of their interaction. It is this synergy that the 1st International Workshop on Data-driven AI (DATAI) aims to explore, delving into the latest research breakthroughs and presenting innovative techniques and methodologies at the forefront of data-driven AI.
This workshop is dedicated to fostering a comprehensive understanding of the intricate relationship between AI technologies and the data they depend on, focusing on the development of high-quality data specifically tailored for AI technologies, with a particular emphasis on large-scale models. Through engaging researchers, developers, and practitioners in rigorous discussions, the workshop seeks to explore sustained advancements, design innovations, and practical applications of data construction techniques that propel the progress of AI technologies forward.
Topics of Interest
Relevant topics include, but are not limited to:
- Data discovery for ML
- Data cleaning & integration for ML
- Labeling quality and ML performance
- Data-efficient solutions for ML training
- LLM-based data cleaning & integration
- Multi-modal data lakes (retrieval-)augmented large langauge models
By fostering a collaborative environment, DATAI aims to inspire a diverse audience of
participants from the
realms of AI and data quality management, facilitating an exchange of ideas that propels the
field toward
groundbreaking developments.
Contact Information
For further inquiries, please contact the chairs through the provided email addresses in the official document[View PDF].
- Hongzhi Wang: wangzh@hit.edu.cn
- Nan Tang: nantang@hkust-gz.edu.cn