Galaxy: Visualize concept associations

Galaxy is an interactive relationship map of the concepts in your text analysis project. It helps you move beyond a list of frequent topics and understand how different ideas connect in the context of your data.

Use Galaxy to answer questions such as:

Journey Insights Hub models concepts using both general language knowledge and the way those concepts are used in your text analysis project. Each concept is represented through many dimensions of meaning.

Galaxy visualizes those multidimensional relationships into an interactive two-dimensional map. Concepts with stronger relationships generally appear closer together, while concepts with weaker relationships appear farther apart.

The map is dynamic. Selecting or moving a concept changes the perspective from which the relationships are displayed, but it does not change the underlying data or associations.

Think of Galaxy as a relationship map rather than a word cloud. A word cloud mainly emphasizes frequency. Galaxy shows both the prominence of concepts and how they relate.

The size of a concept reflects its prevalence. Prevalence measures how unusually common a concept is in your text analysis project compared with language in general. Larger concepts are therefore especially characteristic of the text analysis project.

Concept size should not be interpreted as a direct match count. A concept may appear in many records but remain relatively small if it is also common in everyday language. Another concept may appear in fewer records but appear larger because it is unusually specific to the text analysis project. Use Volume when you need exact record counts or percentages.

The distance between concepts represents the strength of their relationship in the context of the text analysis project. Concepts that appear closer together tend to be used in more similar contexts. They may represent:

Concepts that appear farther apart are less closely related in the text analysis project’s language model. Closeness does not prove that two concepts always appear in the same record, and it does not mean that one concept caused the other. Treat a relationship as a signal to investigate rather than a conclusion on its own.

Related concepts naturally form groups across the map. These groups can reveal the major conversations within the text analysis project without requiring you to define all the topics in advance.

A cluster may contain a central subject together with related actions, attributes, experiences, or outcomes. Concepts between two clusters may help explain how different areas of the conversation overlap. Use clusters to:

When Galaxy is colored by conversation cluster, color distinguishes the groups detected by Journey Insights Hub. These colors correspond with the conversation clusters presented in Highlights.

Cluster boundaries may not always be visually distinct. If the concepts are highly interrelated, the map may appear as one or two dense groups containing several color variations. This indicates that the conversations overlap strongly rather than forming clearly separated topics.

Select a concept to bring its relationships into focus. Galaxy emphasizes the selected concept and its stronger relationships while de-emphasizing concepts with weaker connections.

When you hover over a concept in the Galaxy visualization:

Conceptual matches are limited to the 20 strongest related concepts with an association score of at least 0.5. These may include related variations such as “testing available.” Other related concepts may represent broader contextual relationships, such as “laboratory” or “samples.”

Gray concepts still have relationships of varying strength with the selection and may provide useful context. Their weaker emphasis does not mean they are irrelevant.

The exact colors displayed may change when you select a different coloring option. Use position, emphasis, and the selected color setting together when interpreting the map.

You can also:

Dragging a concept does not create a new association. It changes the projection so you can inspect the existing relationships from another angle.

Default cluster colors can help you become familiar with the major conversations detected in the data. After identifying concepts you want to investigate, apply custom colors to test a different analytical structure.

For example, you might color concepts by:

Custom colors leave the relationships among concepts unchanged. They provide another way to organize and interpret the existing map.

Galaxy is most useful for discovery and exploration. Once you find an interesting relationship:

A strong workflow is to begin with one concept, explore its surrounding language, identify a potentially meaningful relationship and then validate that relationship with the supporting records.