[Paper Review] Generalized Graph Pattern Matching

Pedro Almagro-Blanco, Fernando Sancho Caparrini|arXiv (Cornell University)|Aug 12, 2017

Graph Theory and Algorithms6 references4 citations

TL;DR

This paper introduces Generalized Graph Query (GGQ), a robust logical framework for querying property graphs that extends selection graphs to support complex, cyclical subgraph patterns using powerful predicates and refinements. GGQ enables efficient, automated query construction for knowledge discovery and white-box machine learning, demonstrated through its integration into the GGQ-ID3 algorithm for decision tree learning.

ABSTRACT

Most of the machine learning algorithms are limited to learn from flat data: a recordset with prefixed structure. When learning from a record, these types of algorithms don't take into account other objects even though they are directly connected to it and can provide valuable information for the learning task. In this paper we present the concept of Generalized Graph Query, a query tool over graphs or multi-relational data structures. They are built using the same graph structure as generalized graphs and allow to express powerful relational and non-relational restrictions on this type of data. Also, this paper shows mechanisms to build this kind of queries dynamically and how they can be used to perform bottom-up discovery processes through machine laerning techniques. ----- La mayoría de los algoritmos que aprenden a partir de datos están limitados ya que sólo son capaces de aprender a partir de datos estructurados en forma de tabla en la que cada fila representa un registro y cada columna una propiedad asociada. Estos algoritmos, no tienen en cuenta los atributos de las estructuras con las que un registro dado puede estar relacionado, a pesar de que éstos pueden aportar información útil a la hora de llevar a cabo la tarea de aprendizaje. En este artículo presentamos el concepto de Generalized Graph Query, una herramienta de consulta de patrones en grafos generalizados. Dicha herramienta ha sido construida utilizando la estructura de Grafo Generalizado y permite expresar restricciones relacionales y no relacionales sobre este tipo de estructuras. Además, en este artículo se presentan mecanismos para la construcción automática de este tipo de consultas y se muestra cómo éstas pueden ser utilizadas en procesos de descubrimiento tipo bottom-up a través de técnicas relacionadas con el Aprendizaje Automático.

Motivation & Objective

• To address the lack of standardized, expressive query mechanisms in graph databases for complex pattern detection beyond basic subgraph isomorphism.
• To generalize existing query models like selection graphs to support cyclical patterns, richer predicates, and flexible refinements.
• To provide a foundation for automated, controlled-complexity query generation suitable for white-box machine learning and knowledge discovery.
• To enable efficient construction of complex queries through compositional refinement operations that generate embedded partitions of query structures.

Proposed method

• Proposes GGQ as a logical framework for querying property graphs using generalized graph structures that support nodes, edges, and properties.
• Extends pattern matching beyond isomorphism to include graph simulation and bounded simulation, allowing flexible subgraph matching with constraints.
• Introduces a tree-based evaluation mechanism where each node in the pattern acts as a root, enabling detection of cyclical patterns through path intersections.
• Defines a set of atomic refinement operations that allow constructing complex GGQs from simpler ones, with refinements generating $2^k$ variations for $k$ predicates.
• Employs a declarative query style that supports structural, semantic, exact, and optimal queries, with potential compilation to execution engines like Gremlin.
• Applies GGQ in a proof-of-concept implementation and integrates it into the GGQ-ID3 algorithm for multi-relational decision tree learning.

Experimental results

Research questions

• RQ1 How can graph pattern matching be generalized to support cyclical and complex subgraph patterns beyond simple isomorphism?
• RQ2 What refinement operations enable the systematic, automated construction of complex queries from simpler ones in a controlled-complexity framework?
• RQ3 Can GGQ serve as a robust, extensible query language for knowledge discovery in graph-structured data?
• RQ4 How does GGQ compare in expressiveness and efficiency to existing query languages like SPARQL, Cypher, or Gremlin for complex pattern detection?

Key findings

• GGQ successfully supports the detection of cyclical patterns in graphs by evaluating a tree rooted at every node in the pattern, overcoming limitations of prior projection-based methods.
• The framework enables the construction of complex queries through compositional refinements, with refinement families generating $2^k$ variations for $k$ modified predicates.
• The GGQ-ID3 algorithm, which uses GGQ as decision tree tests, successfully extracts meaningful, complex patterns from graph-structured data for knowledge discovery.
• Refinement sets can be optimized by combining atomic operations (e.g., adding an edge and a property in one step), reducing the number of steps and improving efficiency.
• The framework is extensible to hypergraphs with minor modifications, suggesting a natural path toward generalized hypergraph query support.
• Experimental validation confirms the practical utility of GGQ in real-world knowledge discovery tasks, particularly in multi-relational learning pipelines.