Dilex
A cross-session academic digital library
Unique timeline visualization for assessing past search sessions.
Search results and saved documents are organized by tasks.
Dynamic topic highlighting when search results are scrolled.
Motivation
The idea of this app came to me when I started my grad studies at the University of Regina. Initially, my research topic was not finalized, and my primary task was to read papers about the search behaviour of academic digital library users. While searching for research articles on this subject, I personally experienced the challenges searchers face when diving into unfamiliar domains and assessing research articles without a full understanding of the topic. Organizing information is hard. While we can use our browser history, written notes, or any other digital library tools to keep track of our research works, once we take a break and come back to our task we forget where we had left things. Also, the search history tab of browsers is very messy to refind our previous search activities. My research supervisor and I discussed this problem and decided to develop a digital library search interface to assist searchers in these situations.
Digging deep into the problem
I discovered that such search tasks often require multiple sessions to complete. As a result, searchers frequently return to a search task after a prolonged gap. Many studies suggest that searchers struggle in two main areas when resuming a task after a break. Firstly, they often forget their previous actions and spend significant time reacquainting themselves with their prior work. Secondly, searchers frequently need to re-find information they found useful in previous sessions (e.g., a relevant article or a specific search query that yielded many relevant documents). I found that no current digital library search interface assists academic searchers in both resuming a search task spanning multiple sessions and functioning on both mobile and desktop environments. The second point is crucial because searchers are often not near their primary workstations and rely on mobile devices to continue their academic search tasks.
Developing Dilex
I started developing Dilex using Angular and Bootstrap in the front end and Express and MongoDB at the back end. When searchers issue a query, the Angular app sends that request to Express REST API. The backend then fetches the results from University Resources and returns them back to the app. All other data such as Tasks, Search history, and saved pages are stored in the MOngoDB database. The app has a responsive mobile-first design, created using the Bootstrap CSS framework. The timeline visualization was built using the ECharts js library.
Features
1. Task Dashboard
To help users manage different search projects, I created the concept of separated tasks with 'task cards' on the dashboard page. As a result searchers searchers can easily keep their search history, saved pages and sessions organized. No more digging through the browser's search history! These task cards are sorted with the most recently used at the top. Collectively, this information supports searchers in quickly assessing their past search activities.
Figure 1: Task dashboard Page.
2. Unique Timeline Visualizaion
I also used a novel visualization of representing search sessions with bubbles on a timeline. The search sessions themselves are represented as circles on the timeline. The size and saturation of each circle visualize the number of queries issued and the number of documents saved in that session, respectively. The purpose of this timeline is not to support precise decoding of this information, but instead to convey a summarized story of the past search sessions in a compact way. ECharts library is used to build this visualization.
3. Task Detail Page
Users can re-issue an old search history or they can create a new one. Either way, they will be taken to the search result page. Selecting a task card from the dashboard takes users to the task detail page, providing a detailed view of past search activities. On this page, the timeline is positioned at the top, and the details of each past search session are provided in “session cards” (most recent at the top). The session cards are labelled with the first query issued in that session, and contain all the search history of that task on that session for re-finding a piece of information. Inside the session card, for each search history, the following information is displayed: query title, number of search results returned, number of pages browsed, and number of documents saved. Re-issuing a previous query can be done by selecting the specific query.
Figure 2: Task Detail Page.
4. Search Result & Workspace Page
Users can re-issue an old search history or they can create a new one. Either way, they will be taken to the search result page. I have also used a new scrolling approach on this page for quickly assessing search results without diving deep into the page. A list of topics extracted from all search results is displayed at the top of the page. For each topic, the number of documents that are associated with it is provided. As the searcher scrolls within the result list, the document at the top of the list and its associated topics are highlighted using the same border style. The topics are also sorted dynamically so that only the topics present in the top-selected document are displayed first. This is a unique and new way of highlighting topics as a user scrolls the search result list. The same design and visualization are also applied in the Workspace page where searchers can list results they have saved for this task.
Figure 3: Workspace Page.
User Study
Based on the feedback from my supervisor, I further improved the academic search interface. We then recruited participants for the study from our university, focusing on students who fit the target demographic. Half of the participants used the prototype interface, and the other half used a recreation of the University of Regina digital library search interface. We collected data on their search behaviour, the documents they collected, and their interactions with the system. Analysis of the data revealed that the prototype search interface outperformed the University of Regina digital library search interface in assisting academic searchers to continue their tasks over multiple sessions, both on mobile devices and desktop environments.
I wrote two research papers with my supervisor about the design process and results of our study. They were published in two renowned ACM scientific conferences.