Toward a General Framework for Multimodal Big Data Analysis.

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Multimodal analytics in Big Data architectures involve complex configurations for diverse data types.
• Scalability challenges arise from resource calibration trade-offs across different analytical tasks.

Purpose of the Study:

• To propose a unified methodology for multimodal analytics within a single data processing approach.
• To simplify Big Data architectures and fully leverage parallel processing capabilities.

Main Methods:

• Integrating multiple data sources into a unified knowledge graph (KG).
• Defining ad hoc views on the KG for specific data modalities and rewriting the graph.
• Applying graph embeddings to transform views into a uniform low-dimensional format.
• Utilizing a single machine learning procedure for diverse data modalities.

Main Results:

• Demonstrated reduction in execution costs for multimodal analytics.
• Achieved improved accuracy in data analytics tasks.
• Simplified system architecture through a unified processing approach.

Conclusions:

• The proposed methodology effectively addresses multimodal analytics challenges in Big Data.
• The unified approach enhances efficiency and accuracy, offering a scalable solution for complex data environments.