New in Hume 2.7: Search relevance, improved visualisation, and much more!

· 7 min read

GraphAware is proud to announce the 2.7 release of Hume, our graph-powered insights engine. The release significantly enhances Hume’s knowledge graph visualisation and graph data analysis capabilities. Analysts, data scientists, investigators, and data-savvy business users immediately get the following main benefits:

  • Faster identification of the starting points for analysis through configurable search relevance and improved end-user search experience
  • Speedup in time to insight thanks to the long-awaited, attribute-based relationship styling
  • Increased flexibility for individual analysts brought by automatically computed virtual node and relationship attributes
  • Reduction of dead-end investigation paths by preventing the “hairball” problem through configurable double-click actions
  • Reduction of clutter through the experimental node and relationship grouping functionality
  • Additional opportunities for data enrichment and further processing through the addition of Neo4j Reader component to Orchestra
  • Improved user experience thanks to countless, sometimes unnoticeable but very impactful UX improvements

A new look and more customisable

Graph visualisation and exploration is at the core of Hume’s offer for analysts. This release of Hume improves it enormously:

  • To maximise the space dedicated to the graph, the side panel is now closable with panel titles rotated to fit vertically.
  • Searches and ordering increase the analyst productivity by simplifying the access to relevant information or options.
  • Actions management has been moved to a dedicated panel since they are first-class citizens of Hume’s exploration capabilities. The grouping by scope makes them more clear and better organized.
  • The double click configuration unlocks customization options. Direct users or managers can decide what is the default behavior on double click, thus helping to speed up exploration.

The video illustrates how the new panel behaves and how simple it is to search for and visualise attributes. Furthermore, it shows the rich range of the options available for double click configuration.

The power of enhanced search at your fingertips

In most of the cases, search is how users start their analysis, bringing to the canvas relevant nodes from which they continue their exploration. The more efficient, effective and flexible it is, the faster and more productive the analysis will be. In order to allow search to fulfill all the exploration needs, Hume 2.7 extends its capabilities and customizability in multiple ways:

  • It is now possible to specify the relevance of some classes or some attributes. These settings can be done at the perspective level selecting a specific boost value for the attributes or for all the attributes in a node class.
  • From the search bar, just using the keyboard, it is possible to specify the node type where to search or a specific attribute in the nodes. Autocompletion makes the process easy and misspelling-free.
  • Multiple search criteria can be combined to further refine the filter.
  • Sometimes, the nodes or the relationships returned as a search result are already in the graph currently visualised in the canvas. They are now properly highlighted as new nodes or relationships.

The video shows some examples of how to search by specific attributes or how to change the relevance to customize the results obtained. The example on query combination explains how to specify multiple criteria.

See the invisible

Hume’s main goal is to empower humans with tools that make knowledge workers’ analysis simpler and straightforward. The power of graphs in visualisation is to make patterns immediately recognizable just looking at them. This is where they shine compared with dashboards or tables. Stemming from this idea, in addition to node styles, Hume now introduces link styles. Relationships can be customized in accordance with the values of some attributes. For example, you can change the thickness of the link based on some “weight” attribute. These styles make the graph easier to understand and interpret.

Through the usage of computed attributes, Hume makes it possible to access all contextual information needed to customize style for relationships. Computed attributes are “virtual” properties that can be added, a posteriori, to both nodes and relationships. They are computed using custom queries. One of the most common usages of computed attributes is to customise styles but they are also very valuable as information per se. Sometimes, these attributes are difficult to persist in the graph, for example when these values depend on the structure of the graph and it changes too often. In these circumstances, computed attributes can help seeing what would be hard to see otherwise.

The video reveals all the possible effects users can apply to the relationships. Some of them are very, very cool. In the example of computed attributes an average rating value is “attached” to a node based on the relationships connected to that node.

For those who want to experiment

Starting from this release note, we would like to highlight some of the experimental features we introduce in our major releases. In this way we aim at soliciting your curiosity around some of the functionalities that could be fully fledged features in Hume at some point in the future and, even more importantly, we want to gather feedback and ideas. Hume is a complex ecosystem that requires a lot of product discovery before a new piece will be released, especially for some of the more sophisticated areas. Of course not all experiments are successful. We dropped a lot of ideas that appeared to be amazing at the beginning but then we discovered that they were too hard to be understood and used properly.

Keeping this in mind, our great desire is that you try an experimental feature that we are quite sure you’ll love in the long run: the grouping capability. In short, it allows users to group nodes and, as a consequence, relationships based on some attribute values or some class.

Have a look at the video to see how powerful they could be.

Further relevant improvements

Along with major changes, each release comes with improvements, bug fixes, minor updates and so on. Some of these include:

  • Configurable retry value for Neo4j Writer: It is now possible to specify how many times the Neo4j Writer component in Orchestra will retry a transaction that failed due to concurrency issues.
  • Ability to change System Settings from the application: So far, any change in the system settings required a change in a properties file and the related restart of the entire system. Now, the most critical settings can be edited directly from the application and no restart is needed anymore.
  • Ecosystem Variables: While configuring resources, some fields are filled with the same values again and again. This makes the management of the Hume ecosystem repetitive. To overcome this issue, this new feature allows you to specify these variables once and then re-use multiple times.
  • Lazy load large properties: Some graphs contain long text inside some attributes which makes the loading of the property pane related to such nodes quite tedious and long. Hume 2.7 allows users to specify in the schema that an attribute contains a lot of text (that is the most common use case) and this property will be loaded fully only when the user explicitly asks for it. Otherwise only a snippet will be loaded and displayed.
  • Neo4j Reader: A new, specific component, dedicated to reading from Neo4j has been added. When used to source a workflow, it alone is capable of replacing three previous components: a timer, a cypher query processor and the Neo4j writer. This component starts a read-only session with Neo4j so any change to the graph will be forbidden. This makes the use of this component much safer.
  • Monitoring Orchestra (Prometheus + Grafana): Orchestra handles millions of messages per hour and deals with a lot of data massaging and transformation. It is a crucial element in the process of building and analyzing a knowledge graph. As a consequence, it is vital to monitor it properly and verify that it is performing as expected. We have now tested the integration of this component with Prometheus and Grafana to discover how to gather and visualise as much information as possible about a running Orchestra. The result of this effort is presented here.

Further sources of knowledge

For those interested in learning more about how Hume delivers high value solutions, the following resources provide an overview of the spectrum of possibilities.

If you love art and graphs you cannot miss Tony’s blog posts:

[Blog] Exploring the MET art collections

[Blog] Exploring the MET art collections #2

For who of you have to deal with complex IT operations, the following blog post, from our VP of Engineering, Luanne, is definitely a must read:

[Blog] Insightful IT operations with Hume

If you’d like to read about monitoring in Hume Orchestra, here’s Andrea’s blog post:

[Tech Blog] Hume Orchestra Monitoring

Dr. Alessandro Negro

Research & Development | Neo4j certification

Dr. Alessandro Negro holds a Ph.D. in Computer Science and is a leading authority on graph-based AI and Machine Learning. Dr. Negro is an expert in computer science, graphs, and data science, specialising in natural language processing, recommendation engines, fraud detection, and knowledge graphs. He has written two books on these topics: Graph-Powered Machine Learning (Manning, 2021) and Knowledge Graphs Applied (Manning, estimated publication in 2024) and his expertise is highly sought after within the industry.