The Topic Explorer is a state-of-the-art visualization of AI-derived topic boxes that can be used to view a topical breakdown of search results or to explore topics for further analysis.
You can use the Topic Explorer with or without performing a specific search, depending on your use case. Exploring an entire dataset, by using an empty search, allows users to navigate various topics without having to compile an exhaustive list of keywords or grant numbers. For example, if you are interested in exploring all records related to informatics, you could simply select the topic box labeled "informatics" and explore using this route. You can also use the Topic Explorer after performing a search to see how returned records fall into various topics.
Note that at any point you can maximize the size of the Topic Explorer or results table. Additionally, the topic box sizes are sized logarithmically to maximize the sizes of the AI labels, therefore, the sizes are not directly proportional to the number of records in each box.
There are several ways to view the Topic Explorer and its data:
Search Overview (All Three Datasets) screen
Dataset screen
Accessible Table
Search Overview (All Three Datasets) screen
On the landing page, select All Three Datasets from the drop-down menu. This will bring you to the Search Overview screen, which contains the first preview of all three datasets (Literature, Grant Records, and Clinical Trials) along with the topics captured in each. You can customize the heat mapping to view distributions of various topic box attributes, such as NIH-funded literature or research grants. To interact further with the visualization you will need to select an individual dataset. You can do this by clicking on the visualization or dataset header.
If you have performed an empty search (did not input keywords or other information into the search bar), all records within the entire dataset will be divided into topic boxes. Each topic box has an AI-generated label made up of the most relevant keywords to help you understand its content. Learn more about the AI labels and display levels to see all AI labels/more granular topics. The shading reflects variation in the selected metric across topics, such as percent of literature funded by NIH. Learn more about the heat mapping.
If you have performed a targeted search for specific keywords, grant numbers, publication titles, etc., you will only be viewing topic boxes in which your results have been previously assigned using the AI language model clustering methodology. Note that the topic box assignments are not dynamic, so performing an identical search with the same records will result in the same visualization each time. It is also always important to check the Search Fields that are selected when performing a search with specific text.
To learn more about Search Fields and search strategies, review Search Strategies – Manual query using Search(able) fields.
To learn more about the Search Fields available in each dataset, visit the dataset-specific glossaries:
- Search Options/Glossary (Literature)
- Search Options/Glossary (Grant Records)
- Search Options/Glossary (Clinical trials)
In the example below, the phrase “breast cancer” was searched using the "content" Search Fields of the various datasets:
Dataset screen
To fully utilize all functionalities of the Topic Explorer, you will need to enter the dataset of interest. You can do this by selecting the database on the iSearch Analytics landing page or clicking on the database visualization or header in the Search Overview. Selecting an individual dataset allows you to fully interact with the AI-derived topic boxes and view individual records.
If you have performed a targeted search for specific keywords, grant numbers, publication titles, etc., you will be viewing only records that match the specific criteria entered, a subset of the entire dataset. The Topic Explorer will also be limited to topic boxes in which your results have been previously assigned using the AI language model clustering methodology. Each topic box has AI-generated labels made up of the most relevant keywords to help you understand its content. Note that the granularity of topics you see (broad, intermediate or narrow level topics) is based on the number of results; the system automatically chooses the most appropriate level for you though you can update this by updating the display level. The shading reflects variation in the selected metric across topics, such as percent of literature funded by NIH. Learn more about the heat mapping.
You can learn more about the Dataset/Results view in the user guide or through the walkthrough that appears when you first open a dataset. To close the walkthrough, click the X in the corner of any step. You can revisit it anytime by clicking the Tutorial button at the bottom-left of the screen.
Accessible Table
The default view for the Topic Explorer is the visualization with the heat mapped metrics. An alternative view is an Accessible Table of this data. Whether you are viewing all three datasets or a specific dataset screen, you have the option to view all topics in your results and their corresponding metrics in this table format, using the Accessible Table icon, which can be found beside the Topic Explorer text and table icon above the visualization.
The accessible table will have all topics in the dataset(s), which is different than what is able to be seen in the visualization; the latter in only able to show at most 30 topics in your results. Depending on the display level you are in, along with the number of search results, you may have several pages in the Accessible Table.
One topic you are able to see in the table is called "Unassigned", which includes records that were unable to be appropriately placed into a relevant topic. This topic includes records with limited amounts of text to use for clustering (e.g. short titles and missing abstracts). Learn more about methodology used to define the existing topics and their AI labels. Below are screenshots of what the Accessible Table looks like and functionality included, depending on what view you are in within iSearch.
All Three Datasets
Dataset specific
In the Dataset view, you can select a topic in the visualization or use the checkboxes in the Accessible table to Filter topics or exclude specific topics from your search by selecting all (checkbox located in the top blue header), then deselecting those you are not interested in.