An example network of multiple node types made in Kenelyze, showing movies, actors, directors and writers
Getting started with a graph visualization project initially requires actually getting your source data into a format which structures the data as nodes (data points), links and any properties associated with them. In many cases, you may start off with data formatted as a regular table consisting of columns and rows, which somehow needs to be converted into a graph-readable format. Depending on the structure of your data and the type and complexity of the graph you wish to create, this process can be challenging and time-consuming.
With this in mind, one of Kenelyze’s aims is to significantly speed up the process of going from tabular data to graph-powered insights. The built-in data import process enables users to input any table-based data source, pick the type of network they wish to visualize, and model the nodes, links and their properties right from within the interface. The resulting visualization is then mapped out on screen automatically and ready for further exploration, fine-tuning and sharing. It is also possible to directly export the resulting graph structure to Cypher statements for quick import in graph databases.
Let’s take a closer look at the 6 types of graphs which can be created using Kenelyze:
One Node Type
This type of graph is also called a one-mode or mono-partite network. Here, nodes are always connected to nodes of the same type, with connections based on some shared characteristic between two nodes. As an example, imagine a dataset which contains information on movies and the actors starring in them. We can ask Kenelyze to use the information in the Actors column as the basis for the nodes in the graph (separating values in cells which mention multiple values), and connect nodes whenever they co-star in a movie based on the Title column. Here’s what that looks like when importing data:
Multiple Node Types
This type of graph is also called a multi-mode or multi-partite network, and is probably the most common graph type used in graph databases. An example Actors-Movies network is visualized in Kenelyze in the screenshot above. This type of graph can be used to connect multiple node types (or, columns in your dataset) to each other, where each type gets its own shape, color and any attributes you may want to add. For example, we can connect actors (circles) to the movies they star in (squares), and also connect directors (triangles) to the movies they directed.
This is what that looks like in Kenelyze’s import process:
Predefined Links
In some cases, your data may already be structured as a graph by directly mentioning links between data points in columns. For example, if you have columns named ‘From’ and ‘To’ or ‘Source’ and ‘Target’, you can import this directly into Kenelyze using the Predefined Links option. This type of data structure is also called an Edge List (if cells hold single nodes) or Adjacency List (if cells can hold multiple nodes), and is a common data structure used in the world of graph analysis and visualization:
Adjacency Matrices
Every matrix is actually a graph, with node labels mentioned in the first row and column and links between these nodes when there is a value at their intersection in the matrix. Kenelyze can directly import data formatted as a matrix and visualize the resulting graph on the fly:
Text Similarity
This type of network is based on Natural Language Processing of the values in a column of your choice in a dataset, and is an excellent choice if you’re looking to cluster textual data such as documents, summaries or customer reviews based on their similarity. Kenelyze converts the chosen text into TF-IDF vector representations, and automatically plots a graph based on the similarity scores calculated between documents. It is possible to label the nodes based on any other column (for example, the title of a document) to make better sense of the network. Here’s what this looks like in the data import process:
Term Co-Occurrences
A second network type based on text data focuses on generating visualizations of co-occurring terms in a column of your interest. Similar to word clouds but with the added value of being able to view connections, this is a valuable type of network if you quickly wish to get an overview of themes or topics in a dataset:
Populating Graph Databases With Data
A key advantage of being able to quickly move from tables to graphs is that this allows one to very quickly populate graph databases with new datasets. When creating a network in Kenelyze, it is possible to export the nodes, links and any properties they hold to a file which contains all the Cypher statements necessary for importing the data directly into Cypher-compatible databases such as Neo4j, Memgraph, RedisGraph, AnzoGraph and ONgDB.
Interested in a demonstration and trying this out with your own data? Schedule a demo now.
