1 ’s System Design
Developed using modern web technologies, ’s interface provides users with new functionalities beneficial for literature sensemaking (e.g., incremental exploration, live data connection to Semantic Scholar) while inheriting the core graph visualization capabilities of Argo Lite [4Li2020ArgoLO] (e.g., React for Blueprint for UI, MobX state management, Three.js WebGL graph rendering).
Displaying Paper Relations. ’s main view displays a citation network as a graph of papers (nodes) and citations (edges). Users can add any paper indexed by Semantic Scholar [5Lo2020S2ORCTS] (https://www.semanticscholar.org) by entering the paper’s unique CorpusID via the Papers Menu (: Interactive Visual Exploration of Literature in BrowsersB). From there, the user can branch out from currently visualized papers by exploring their citations and references or input new, potentially unrelated, papers. Directed edges point from a citing paper to the cited; edge directions can be toggled. A variety of visual properties of the nodes, edges, and node labels can be customized via the Customization Panel (: Interactive Visual Exploration of Literature in BrowsersE). For example, nodes can be colored and sized based on paper attributes such as citation count, with the ranges of both being user-adjustable.
Incremental Exploration. Making sense of literature through incremental paper exploration or associative browsing is a common, effective method for researchers to discover relevant work [Kairam2015RefineryVE, 3Chau2011ApoloMS]. Specifically, they may start with a handful of (familiar) papers, then gradually build their network of related papers through exploring the references and citations of their growing network. supports this literature sensemaking strategy; users can iteratively add 5 reference or citation papers for a paper333Semantic Scholar supports 100 requests per 5 minutes per IP address via its Exploration Dropdown (Fig : Interactive Visual Exploration of Literature in BrowsersD). The paper is connected to the added papers by directed edges that encode the citation relationships previously described, and the added papers are aligned vertically to improve label readability. Network data, such as the degree, pagerank, size, and color of existing nodes, are updated accordingly.
Learning More About Each Paper. The Paper Information Panel (Fig : Interactive Visual Exploration of Literature in BrowsersC) of shows essential paper information for a highlighted or hovered-over node, e.g., paper title, abstract, authors, citation count, publication venue, publication year, and the URL to its Semantic Scholar page (for paper PDF, figures, etc.).
Saving and Sharing Explorations Across Platforms. allows users to save and share their literature exploration as a snapshot JSON file that stores the information of all explored papers, their connections, and visualization customizations via the Graph Menu (Fig : Interactive Visual Exploration of Literature in BrowsersA).
Users have two options when saving and sharing their exploration snapshots: locally on their device or in the cloud. If users opt for the first option, they will download the JSON file and then upload it to whenever they wish to resume their sensemaking process. Or, users can save their snapshots to ’s server for free, enabling them to share their literature network as a custom URL, bypassing the need to send JSON files. Saving the snapshot to ’s servers also grants the ability to embed the literature network as either an HTML IFrame or a Jupyter Notebook IFrame. runs on all modern web browsers (e.g., Chrome, Safari, Firefox, Edge), broadening its access.
2 Conclusion and Ongoing Work.
is an open-source, web-based visualization tool for interactive exploration of literature and easy sharing of exploration results. queries and visualizes Semantic Scholar’s live data, enabling users to generate personalized literature exploration results in real-time through flexible, incremental exploration. Anyone can access our tool using their web browser on desktops and mobile devices without the need for any software installation.
We plan to improve the user experience of by enabling users to directly search and show paper results in , bypassing the current need for locating the paper’s CorpusID. This tighter connection with Semantic Scholar would allow us to enhance the usability and user-friendliness of our visualization tool. Additionally, we plan to improve the ordering of adding cited papers and referenced papers by implementing the ability to add them based on metrics such as citation count, recency, or relevance. These improvements would add more flexibility to the sensemaking process as users can further customize the way they explore literature. We also plan to evaluate through user studies with both beginning and seasoned researchers to evaluate ’s usability and its ability to help with literature sensemaking in the long term. As improves its functionalities, we look forward to more students, researchers, and practitioners adopting to discover relevant research and make sense of the literature across disciplines.