LINCS Context Explorer

Using named entity recognition, users can scan a web page connecting named entities found in the web page to data points (otherwise known as ‘entities’) found within LINCS, including people, organizations, places, and artistic works. From there, users can explore entities' relationships and learn how they’re connected through an overlayed node-edge graph and an informational side panel.

Using this information, academic researchers and people interested in cultural knowledge can discover new connections and relationships.

The project first began as a proof-of-concept, developed in Summer 2021. As a part of LINCS' official launch in April 2023, we sought to build an MVP version of the tool. As the sole designer on this project, my task was to guide the direction of the experience, in addition to designing the interface.

To initiate the design process, I first conducted a stakeholder meeting to establish the project's goals and requirements. In this meeting, I discussed usability issues found in existing LINCS tools, focusing on the issue of technical language. Past studies on previous tools, found that users with limited knowledge of linked data had difficulty creating value out of LINCS data due to the complexity of language. As such, research questions could not be answered and interesting discoveries could not be made.

With this in mind, we established some high-level use cases:

Based on the requirements and outlined use cases, it was clear that we had to focus our tool for casual users. As defined from previous research, casual users comprise of academics, scholars, and the general public that have an interest in exploring cultural information with no knowledge of linked data. With a broad user profile, I opted to focus more on specific shared behaviours between these different users rather than creating individual personas.

Using pre-existing user stories and data gathered from previous user research, I outlined behaviours relevant to accessing linked data which, in turn might affect the usage of the tool.

The main goal I wanted to accomplish with the design was to guide the user to information they value and understand. This included:

To strategize the experience I wanted to create, I outlined a high-level user flow that focused on three stages: scanning, matching, and exploring.

During this process, I conducted two usability tests to validate and iterate on my designs.

Scanning

While designing this stage, a significant part of the effort went into the microinteraction. To initiate my designs, I first began ideating on various ways to initiate a scan and select a scanned and highlighted word.

Through initial discussion with the team, we decided that users would hover to view the word's LINCS matches. However, in my initial usability test, I found that all participants continuously attempted to click on highlighted words, causing issues when the word was a hyperlink.

Another challenge was that we couldn't just change the interaction to a click, as we didn't want highlights to disable a webpage's hyperlink functionality. Therefore, I found a compromise between clicking vs. hovering. I modified the interaction to display a clickable selection button over the highlight on hover , to hopefully instill a behaviour for hovering first.

To discover if this worked, I ran a new usability test to find out whether participants could learn to not click the highlight. From a small sample group, I found that 3/10 users were able to either discover the interaction on their first try or recover from their error on a subsequent attempt. Although not an ideal number, it was still an improvement from the previous result.

Matching

After users select a highlight, they then are presented with a list of matches that are found within LINCS data. LINCS has over 50,000 entities, all with varying degrees of information — as such, one important aspect was to assist users into selecting results that had potential information value. As such, I came up with an indicator that displayed the number of connections. Results with higher counts would often have greater information value over results with lesser counts.

Exploration

In this stage, we wanted to display what relationships our matched entities had with other LINCS entities. A key part of designing this stage was determining how to represent this relationship visually. Initially, the proof-of-concept contained all interactions and information within a node-edge graph, causing issues with understandability due to a lack of visual differentiation. In addition, I knew that only containing a graph could create challenges for screen-reader accessibility. To address this, I created parallel representations through both an overlayed node-edge graph and a new side panel.

First, I looked to redesign the graph, sketching out ideas that focused on better clarity of system state to indicate what entity the user is currently exploring. By placing selected entities in a central pill, users would have a better idea of what relationship they're analyzing.

From there, I began work on designing the side panel. To communicate a relationship between two entities, and their resulting relations, I divided the panel into those respective sections, to establish consistency with the graph. To display the relationship between entities, I iterated on various nested designs.

After multiple rounds of design feedback and content revisions, we decided on the design below. In the first usability test, I found that all 5 participants understood the visual format, being able to identify the entities' relationship and relations.

I had the opportunity to design the content as well. The side panel contained a dropdown that information specific to a selected entity. However, usability tests found that participants didn't understand this information, causing them to ignore the section.

Thus, I helped created a more informative summary section for the side panel dropdown. By auditing existing datasets in LINCS, I identified common data points represented in LINCS entity types (i.e categories like people, places, and works) to create a summary framework.

Thanks to the hard work of the team, the Context Explorer was presented and featured at various academic conferences. At Congress 2023 of the Humanities and Social Sciences, as part of the Canadian Society for Digital Humanities, I held a demo session and at Making Links 2023, was also a part of a UX panel discussion talking about the tool.

This was one of the first projects where I had the opportunity to take ownership of the design to development. Throughout this process, I learned a ton about communicating with developers, from handing off designs, understanding how interface components are created, and how data is used.