Scalable DB+IR Technology: Processing Probabilistic Datalog with HySpirit.

Area of Science:

• Computer Science
• Artificial Intelligence
• Database Systems

Background:

• Probabilistic Datalog (PDatalog) offers a logical, rule-based approach to Information Retrieval (IR).
• Existing PDatalog implementations face challenges in real-world application scalability and feature integration.

Purpose of the Study:

• To present key features of the HySpirit system for scaling Probabilistic Datalog execution.
• To demonstrate how HySpirit addresses requirements for real-scale data management and analysis applications.

Main Methods:

• Integration of probability estimation capabilities within PDatalog.
• Inclusion of fuzzy-like predicates for modeling vague predicates.
• Implementation of probabilistic relational indexes and parallel/distributed processing for scalability.

Main Results:

• HySpirit enables the expression of probability estimation in PDatalog programs.
• Vague predicates are effectively modeled using fuzzy-like predicates.
• Scalable execution of PDatalog is achieved through probabilistic relational indexes and distributed computing.

Conclusions:

• The HySpirit system provides a consolidated approach to making PDatalog applicable for real-scale data management.
• HySpirit's features address critical requirements for probabilistic reasoning in complex data environments.