Main Components

A. The search panel is used to search for documents within the database. Search using the keyword search or the neighbors search. Keyword search will display a list of matching documents underneath the search bar in the panel, and show green points on the corpus map representative of those search hits. Enter in your own abstract and click neighbors search to input your own abstract and compare it against other papers in the database.

B. The corpus map is a tool to navigate through documents in the database. It displays all documents in the database as a density heat map, and explicitly shows documents that are part of the search, selection, or neighborhood. It is a 2D projection of whatever embedding is selected. Generally, this can be interpreted that if points are closer they might be more similar, however the dimensionality reduction can distort this.

C. The scatterplot displays an expanded view of regions. The corpus map is organized into cells, where each cell is a region of the map. Clicking on one of the cells expands it in the scatterplot, where you can see the region in better clarity along with other papers that aren’t explicitly shown as a point in the corpus map.

D. The regional matrix allows for easy viewing of what salient words have the highest frequencies between all the documents within a region, with sorting metrics to change what order the words are sorted in.

E. The regional list allows for systematic exploration of the region with multiple sorting criteria.

F. The neighborhood matrix is used to compare salient words between similar documents, displaying frequencies of each salient word with each document in the neighborhood.

G. Selected document view lets you view the paper more in depth and also interact with various parts of the paper. Click “yellow highlight on” to highlight the salient words of the paper when one paper is selected, or the common salient words when two papers are selected. Click “add to favorites” to add the paper to the favorites list, which can be accessed using the top right button.

H. The neighborhood list displays a list of similar documents which make up its neighborhood. These are sorted using vector spaces which can be changed using the dropdown.

I. The radial plot can be used to compare the relationship of the top chosen number of similar documents to the first selected document. The closer points appear to the center point, the more similar they are according to the vector space chosen. Advanced settings are available.