The Project Explorer provides a home-base for the discovery stage of your analysis, with various tools designed to help you explore and uncover key themes in your dataset. Here we provide a brief introduction to all the major features and functions in the Analytics program from concepts to topics, filters to exports, setting the stage for more sophisticated analytical techniques.
The majority of the examples below come from a project built from publicly available Beer Advocate data, chosen because it provides a good variety of easily understandable concepts and good quality metadata to illustrate key tasks for this guide.
The Concept Cloud
The first thing you notice when you open the Project Explorer is the Concept Cloud, a colorful map of the 500 most relevant concepts in your dataset.
- The larger a concept is, the more relevant it is in the dataset
- Related concepts are clustered together and reveal discussion themes
- Clicking on a concept in the cloud reveals a heat map of dark blue highly-related concepts, light blue less-related concepts, and gray unrelated concepts
- The colors in the Concept Cloud indicate how related a concept is to your Topics (see Topics, below)
- Filtering the dataset results in a reshaped Concept Cloud (see the Filter Tool, below)
Analysis is fueled by deep exploration of important concepts in the dataset and determining their associations with one another. There are several ways to explore concepts:
Click on a concept in the Concept Cloud.
Search for a specific concept in the Concept Search Box.
Review the list of top concepts in the Concept Detail Pane.
Read document excerpts in the bottom half of the Concept Detail Pane for a sense of how a concept is being used in context. Within the document excerpts, Exact Matches are indicated with a solid underline and Conceptual Matches with a dashed underline.
Luminoso automatically suggests a set of potentially interesting concepts when you first explore your project, which are grouped into 7 clusters of 4 to provide a broad overview of the key concepts in the project. You can choose to save all, none, or a subset of them as topics for further exploration.
If you identify a concept that you would like to explore further as part of your analysis, you can save it as a topic by clicking the blue plus sign next to the concept's name in the Concept Search Box, at which point it will be added to the list of topics in the Topic Panel on the left. Concepts that have already been saved as topics will appear with a colored dot instead of the blue plus sign.
The Filter Tool
The Filter Tool in the left-hand panel offers an opportunity to drill down into subsets of data that you can define dynamically. When you select a filter, your Concept Cloud and selected documents will adjust to your selection. You can filter for metadata unions by selecting more than one value within a metadata field. You can also filter for metadata intersections by selecting for values in different fields.
String metadata can be selected via checkbox (for fields with 30 or fewer values), or via search box (for fields with more than 30 values).
Numeric metadata can be filtered by a range of values.
You can also filter by date ranges.
It can be helpful to have access to the raw data to tailor your analysis more specifically to your use case.
You can export the following data:
- Match counts (for top concepts or topics): For each concept, get the number of documents that contain the exact concept and the number of documents that contain a related concept. Additionally, the results may be broken down across values of a metadata field.
- Topic-filter association scores: For each topic, get the average association between the topic and each selected filter.
- Topic-topic association scores: Get the association between all topics.