Portolan is an AI-powered, graph-based file management platform designed to streamline research workflows. It helps researchers identify, evaluate, and visualize their research.
Designing for research:
Literature reviews
Research is cumulative. Discoveries build upon previous discoveries. Robust literature reviews are crucial to developing sound and valuable research.
This overview of the product will cover design experiences that relate to this critical part of the research process.
I break the experience down to three components:
1. Visualizing Research
2. Identifying Research
3. Evaluating Research
1. Visualizing research
Literature reviews are overwhelming, often leading to a loss of context. To regain that context, researchers can visualize their research in graph and list views, allowing them to see the connections and overall structure of their findings.
Overview of the graph
Portolan's core architecture is built around the graph data structure. The graph is divided into two components: vertices termed 'Ports,' correspond to individual sources, and edges, termed 'Rhumbs,' represent the connections between these sources.
Node
Designing the node, I wanted to focus on displaying the most critical information at a glance. I had to weigh information and necessity/simplicity.
This led me to design a node that shows the title of the resource and the author. On the top-left is the file status and more options (see below).
If the user wants to see more information such as the publisher, the volume and issue number and the date of publication, they cab toggle it in the settings on the header.
Edge
The figure illustrates how nodes are connected to each other by edges. Each node has a center which edges are attached to. This bi-directional linking provides the user with a richer context for each node.
The arrows function to point the source to where it is being cited. The user can toggle on or off the arrows in the settings tab in the header.
Indicating file status
Portolan is a cloud-based service that allows researchers to upload files and link them to the graph as nodes.
A node has two modes: unlinked and linked. When the node is linked to a file (or external page), the researcher will be able to view the document either within Portolan or outside.
To denote the status, an unlinked node will display an open dashed circle on the top-left corner. A linked node is denoted with a cloud or open link glyph to indicate file location.
Designing the list view
In addition to visualizing the research with the graph, we also adopted a list-view to cater to linear cognitive styles. I translated the information tree into a list-based design system.
The design here parallels the information system of the graph. The Rhumbs are indicated on the left of the list to show connections and the Port hierarchy is shown with the indicators on its right.
2. Identifying research
What to review and what not to review? Some research is more pertinent than others. For example, a short research paper will often have as many as 10 references and many times, you don't need to review all of them.
In addition to this, another challenge for researchers is gathering up the references within a research paper or book. The problem is the location of these references vary between document types. We had to design an experience that allows the researcher full control over the kinds of references they want to retrieve and review.
Importing references into the project: Drag selection tool
The challenge of navigating extensive literature reviews required the design of an interface that prioritizes accurate source extraction. To facilitate this, a drag selection tool allows researchers to pinpoint and extract specific reference segments which can then be automatically integrated into their research graph by preprocessing the selection with Tesseract and processing the result with Gemini into its components (e.g. author, title, publisher, volume, date).
In addition to this, researchers can also search the whole document for citations, select pages, or by page number. These features give the researcher full control over what references they want to collate regardless of the variations in document type (e.g. books, book chapters, research papers, articles).
Using Gemini to search for sources online
Portolan offers users multiple ways to connect sources to their nodes. Users can import directly from their Portolan library or local files, or manually input external URLs or DOIs.
Another option which I want to highlight is using an integrated Gemini API for searching the precise source online. The search retrieves the source and provides access information which enables direct file upload to cloud and node linking.
3. Evaluating research
Portolan is much more than a research management platform. Researchers can also use Portolan to evaluate the source materials linked to the project.
Evaluate and visualize in one view: Quick view
Researchers can preview the content linked to the node in Quick View. This allows researchers to quickly evaluate the information within the context of the research project. As a result, we eliminate the constant need for context switching which significantly reduces cognitive load and creates a more productive research environment.
From Quick View, they scan for references, download the document to their local computer, as well as open the file for more comprehensive reading and editing capabilities.