Journey Insights Hub concepts and terminology

Journey Insights Hub analyzes natural language text and produces insights from open-ended feedback.

It uses natural language understanding to learn how words and phrases are used within a dataset. It identifies the concepts that are most meaningful to a text analysis project, along with their exact and conceptual matches.

Journey Insights Hub uses proprietary natural language understanding and transfer learning technology to analyze unstructured text. The technology is designed to reduce the labeled data, setup time, and extensive training traditionally required to understand domain-specific terminology.

It combines:

These sources are used to create a semantic space of word embeddings tuned to the project data. The background semantic model provides a general understanding of common language. Transfer learning then adapts that knowledge to the terminology, meanings, and relationships found within the dataset.

This allows Journey Insights Hub to learn project-specific language from context without requiring teams to manually label thousands of examples or define every topic in advance.

Highlights provides a quick view of high-level insights from your text analysis project.

Use Highlights to familiarize yourself with the data, identify important areas of interest, and begin exploring the questions that Journey Insights Hub can help you answer.

Use Volume to understand how concepts are represented throughout your dataset.

Volume automatically measures how frequently concepts are referenced across the text analysis project using the exact and conceptual matches identified by QuickLearn.

Galaxy displays strong conceptual relationships within the context of your entire text analysis project. The size of a concept in the visualization indicates how relevant it is to the text analysis project. Related concepts are positioned near one another and cluster together by theme. You can also use Galaxy to save and manage concepts.

Drivers identifies concepts that are prevalent and have a significant impact on a measurable aspect of your data.

For example, imagine a customer satisfaction survey text analysis project in which the measurable score is an overall satisfaction rating. The concept “difficult to navigate” might be identified as a negative driver.

A driver’s impact represents the difference between:

If “difficult to navigate” has an impact of -1.2, records containing that concept or related language have an average satisfaction score that is 1.2 points lower than the overall text analysis project average.

Drivers identifies associations with a measurable outcome. It shows relationships between concepts and score differences without demonstrating that a concept caused those differences.

Use Sentiment to evaluate whether concepts are expressed positively, negatively, or neutrally based on their context. Sentiment provides a detailed view of how customers, employees, or other audiences feel about different aspects of an experience, product, or service.

A record is a single unstructured natural language text sample, along with any optional metadata, that you upload to Journey Insights Hub. In a CSV file, each row usually represents one record. Most Journey Insights Hub text analysis projects contain hundreds or thousands of records.

Metadata is any additional information associated with the unstructured text in a record.

Metadata may include:

After uploading metadata, you can use it to sort, compare and filter your data to investigate specific insights.

A Text Analysis Project is the basic unit of analysis in Journey Insights Hub.

Each text analysis project contains a collection of unstructured text records and their associated metadata. Because each is modeled according to the words and phrases used in its records, its concepts, relationships and metrics are unique.

Filters allow you to use metadata to focus on a specific subset of records.

For example, you might filter a customer feedback text analysis project to view only:

This helps you investigate how findings differ across audiences, time periods, or experiences.

A concept is a meaningful word or phrase that appears in your dataset. A concept may be a single word, such as “confusing,” or a phrase, such as “long wait time” or “not easy to use.”

A theme is a broader area of conversation represented by a group of related concepts. Themes help organize individual concepts into topics that align with an organization’s questions or analytical goals. They may be identified during exploration, created manually, or generated through AI-generated Themes.

A concept list is a saved collection of concepts organized around a theme, business question, or area of interest.

Concept lists can be reviewed, edited, assigned colors, and reused across Volume, Galaxy, Drivers, Sentiment, and Dashboard. Renaming a saved concept changes its display name but does not change how the concept interacts with the text analysis project.

An active concept list is the concept list currently selected for use in the analysis. Activating a list allows you to display and compare its concepts within the current visualization. Always confirm which concept list is active before interpreting or sharing a result.

An advanced concept is a customized concept created to represent a more precise meaning than an individual concept provides.

Advanced concepts can combine related words or concepts or exclude an unwanted meaning. Review the matching records after creating an advanced concept to confirm that it includes the intended language.

A conversation cluster is an automatically detected group of concepts with strong relationships within the text analysis project.

Conversation clusters help reveal how broader conversations are structured in the data. Unlike concept lists, they are generated from the relationships in the text analysis project rather than manually curated for ongoing analysis.

An exact match is a word or phrase that is identical, or nearly identical in form, to the concept you selected. For example, if you analyze the concept “login,” exact matches might include “login,” “Login,” and “logins.” Closely related concepts such as “sign in,” “password,” or “account access” would not be considered exact matches.

A conceptual match is a word or phrase that is meaningfully related to a selected concept but is not an exact match. Conceptual matching helps you identify relevant feedback that may be missed when searching only for exact wording.

For example, if you select the concept “difficult to use,” conceptual matches might include:

An association score measures how strongly two concepts are related within a text analysis project.

Scores range from -1.00 to 1.00, representing the weakest and strongest possible relationships. A concept has an association score of 1.00 with itself. A score of 0 indicates that the two concepts are no more closely associated within the text analysis project than they would be by chance across the language as a whole.

A concept is prevalent when it appears more frequently in a text analysis project than it does in the language as a whole.

For example, in a text analysis project containing feedback about a mobile application, the phrase “verification” may appear far more frequently than it typically does in general English. This indicates that the concept is particularly relevant to the text analysis project, even if it is not one of the most frequently mentioned concepts overall.