April 2024 Summaries

The first version of the ISO GQL standard, a new query language for property graph databases, has been published on April 12, 2024. Developed by an international team of academia and industry experts, GQL aims to become the de facto standard for property graph databases, addressing the growing demand for such databases. The publication of this elegant standard marks a significant milestone in the history of databases. GQL uses pattern matching syntax to declaratively ask queries against graph databases, similar to SQL for relational databases. It supports two syntax flavors: one is Cypher and the other is SQL. GQL also supports a Linux file system-style directory hierarchy to host graph schemas and their catalog objects. The standard is available for purchase and study.

The first version of the ISO GQL standard, a significant milestone in database technology, was published on April 12, 2024. GQL is an ISO standard defined for property graph databases and aims to become the de facto query language standard for such databases. It uses pattern matching syntax to declaratively ask queries against graph databases, similar to SQL for relational databases. The publication of GQL establishes a solid foundation and drives the prosperity of graph databases in the coming years. TigerGraph is actively implementing GQL and contributing to its development.

In this article, the author discusses the importance of optimizing data element definitions to work with Language Learning Models (LLMs) in knowledge graphs. The ISO/IEC 11179 Metadata Registry (MDR) guidelines are used as a basis for creating precise, concise, distinct, noncircular, and unencumbered definitions that can be easily understood by LLMs. These definitions should focus on the semantics or meaning of data elements while also considering their representation in terms of data types. The article provides examples of vertex attributes and enumerated values with clear definitions following ISO 11179 standards.

Graph embeddings are becoming increasingly important in Enterprise Knowledge Graph (EKG) strategy as they enable quick finding of similar items in large billion-vertex EKGs. They aid real-time similarity ranking functions in EKG and can be used for recommendation, next best action, and cohort building. Graph embeddings are small data structures that absorb a great deal of information about each item in an EKG and compress it into compact and easy to compare structures. They enable real-time similarity calculations that can be used to classify items in the graph and make real-time recommendations to users. The process of creating a new embedding vector is called "encoding" or "encoding a vertex", while the process of regenerating a vertex from the embedding is called "decoding" or generating a vertex. Graph embeddings work with other graph algorithms, such as clustering or classification, and can be used to increase the performance and quality of these other algorithms.