Passages: Interacting with Text Across Documents

Title of the Paper

Passages: Interacting with Text Across Documents

Bibliographic Information

• Field of Study: Human-Computer Interaction and Knowledge Workflow Optimization
• Keywords: Knowledge work, document management, provenance tracking, understanding construction, active reading, materialization

Research Background and Problem Statement

• Identified Problems or Challenges: Contemporary knowledge workers face complex tasks when dealing with large volumes of related documents, including challenges in searching, collecting, annotating, organizing, writing, and reviewing. Additionally, they must manually maintain document provenance and interrelations, which is time-consuming and fragile.
• Significance of the Problem: Knowledge work is a core task across multiple industries, and the tools and workflows used directly impact work efficiency and quality. Specifically, patent examiners and scientific researchers must handle highly complex document sets, yet existing tools fail to effectively support seamless and integrated workflows.
• Research Motivation and Related Work:
  • Current software tools often create information silos, making information difficult to reuse and integrate into a unified view.
  • Research in academic literature on knowledge work highlights that active reading, understanding construction, and provenance management are core activities for enhancing knowledge work, but these activities lack cross-tool collaborative support.
  • The authors selected patent examiners and scientists as "extreme user" groups of knowledge workers to study the shortcomings of existing tools and the challenges of their work practices.

Proposed Solution

• Proposed Solution:
  • Introduced the concept of "Passages": materializing text fragments as interactive objects with provenance, annotations, and tags that can be transferred and reused across multiple tools.
  • Developed a prototype consisting of six application modules: Viewer, Searcher, Table, Canvas, Editor, and Reader, all of which share a Passages sidebar to support cross-application sharing.
• Innovative Aspects:
  • Passages addresses the problem of data isolation between tools through materialization and reuse design principles.
  • Enables seamless transitions between multiple document-related tasks while retaining text fragment provenance.
  • Enhances existing tools with stronger cross-process collaboration capabilities rather than creating new isolated applications.
• Implementation Steps and Key Technologies:
  • Developed the front-end and back-end using VueJS and NodeJS, with various databases (e.g., NeDB and PostgreSQL) to implement full-text search across documents.
  • Each module supports materialization, sharing, and drag-and-drop functionality for text selection, with the sidebar recording the originating document and specific location of fragments.
  • Provides an editor to quickly export table or canvas content as structured text, facilitating direct reuse during writing.

Research Outcomes

• Specific Outcomes:
  • By materializing text fragments, Passages allows users to easily switch between browsing, organizing, searching, writing, and disseminating knowledge.
  • The Reader module in the prototype tool quickly displays document content and verifies provenance, improving efficiency and reducing redundant work.
• Advantages Compared to Existing Tools:
  • Unlike traditional single-module tools, Passages' interconnected modules include provenance tracking functionality.
  • Even in complex and diverse knowledge work scenarios, users can flexibly adapt and achieve reusable information fragments.
• Experimental or Evaluation Results:
  • Patent Examiner Testing: Participants found Passages elegant and powerful, particularly in document management and provenance verification.
  • Scientist Testing: Scientists quickly adapted to Passages and found the table module especially helpful for identifying patterns and insights across documents.
  • Passages demonstrated superior performance in reducing cognitive load, improving task proficiency, and increasing success rates.
• Limitations and Future Directions:
  • The current prototype does not fully support other data formats (e.g., video); future work will expand to include non-text content.
  • The Editor's functionality limits user experience and requires further optimization, including advanced features such as rich formatting and annotation pushing.
  • Broader integration of Passages with existing commercial and open-source systems is planned, along with API support.

Conclusion

Passages presents a solution to the complexities of knowledge work, enabling users to maintain provenance and achieve seamless cross-tool interaction for complex document sets. Its design principles and modular functionality provide robust support for improving the efficiency and experience of contemporary knowledge workers, while opening possibilities for applications across multiple domains.