After the text has been processed, MST uses an additional technology to examine the relationships between the data.
Graph databases have a quite different data model than a typical SQL system. A graph database is one that, as a central part of its data model is the ability to store, process and query relationships. The more traditional and common relational SQL database can compute these relationships via queries, but it is expensive and complex in computing terms. By contrast, the relationships are stored in a graph database as persistent connections, allowing unbelievable speed. Graph databases are used, for example, by Amazon for product suggestions, by banks for monitoring credit card transactions, and now, in scientific applications like our ARE4 system.
Regardless of the size of the data, ARE4 excels at managing complex queries because the “heavy lifting” is already done by the relationships in the graph database. By using just a pattern and a set of starting points, our system explores the larger “neighborhood” around those beginning points, then aggregates that information from the larger dataset.
These relationships provide semantically relevant connections between data points. Often, relationships have quantitative properties, like sizes, distances, strengths or time. Because these relationships are stored so efficiently, essentially unlimited relationships are possible between data points with no loss in performance.
Medical Search Technologies uses the state-of-the-art Neo4J™ graph database, which is recognized as the world’s premier graph database engine. It’s an efficient, scalable and powerful product and completely focused on data relationships.
The graph database is an essential component for making our ARE4 natural language processing system understand. Exploring these relationships are a key to understanding and discovery of the underlying concepts in biomarker research.