CorpSum: Towards an Enabling Tool-Design for Language Researchers to Explore, Analyze and Visualize Corpora

Title of the Paper

CorpSum: A Visualization Analysis Tool Design for Linguistic Researchers

Bibliographic Information

• Subject Area: Human-Computer Interaction, Linguistic Visualization Tool Design, and User Interface Research
• Keywords: Linguistics, Text Analysis, Information Visualization, Multi-query, User Interface Design, Corpus Research, Interaction Design, System Usability, Innovative Tools

Research Background and Issues

• Identified Problems or Challenges:
  • Linguistic research relies on large annotated corpora to validate language hypotheses, but existing tools are predominantly based on the traditional "keyword-in-context" (KWIC) model with limited support.
  • Researchers frequently switch between tools (e.g., SketchEngine and Excel), leading to inefficiency and high cognitive load.
  • Current tools lack support for multidimensional comparative analysis or integrated visualization, failing to address user needs comprehensively.
• Significance:
  • Supporting linguistic researchers in conducting efficient, data-driven linguistic studies will enhance the speed and accuracy of academic research.
  • Visualization analysis facilitates intuitive exploration of corpus patterns, reducing the learning curve.
• Research Motivation and Related Work:
  • Existing corpus analysis tools (e.g., AntConc, SketchEngine) generally lack multi-interaction interfaces.
  • These tools have limited capabilities in analyzing temporal changes, semantic shifts, and regional language feature comparisons.
  • To address these limitations, a new paradigm for user interface design is proposed.

Solution

• Method or Solution:
  • Developed an interactive visualization tool called CorpSum, which supports multidimensional queries and interactive analysis within a single interface.
  • Designed a novel layout principle of "overview, zoom, filter, details-on-demand," moving away from traditional text snippet-only interfaces (KWIC).
  • Provided a highly flexible query interface that supports simultaneous comparison of multiple queries, creation of sub-corpora, and export of visualization results.
• Innovations:
  • Integrated corpus processing engines with visualization design to offer a comprehensive analysis platform for linguistic researchers.
  • Broke away from the traditional "co-occurrence analysis-driven design paradigm" by introducing a single-page, multi-view interactive layout, reducing cognitive load.
  • Adopted a user-centered design process, including iterative prototype development and close collaboration with domain linguists.
• Implementation Steps and Key Technologies:
  1. Requirement Analysis: Conducted interviews with 10 linguistic researchers to identify task requirements.
  2. Conceptual Design: Created low-fidelity prototypes and gathered feedback through expert reviews.
  3. High-Fidelity Prototype Development: Built interactive regions using tools like Tableau and tested interface functionalities.
  4. Final Tool Development: Developed the final web application based on VueJS, optimizing performance and interface consistency.
  5. Deployment and User Testing: Validated the tool's effectiveness in corpus analysis.

Research Outcomes

• Specific Achievements:
  • Compared to traditional tools, CorpSum significantly accelerates the query and analysis process (task completion time improved by 3.18 times).
  • Achieved a System Usability Scale (SUS) score of 88.75/100, far surpassing SketchEngine's score of 45.63.
  • Two real-world use cases demonstrate that CorpSum is widely applicable for regional language studies, historical lexicon analysis, and social discourse research.
• Advantages Over Existing Solutions:
  • Supports parallel analysis of multiple queries, eliminating the cumbersome process of combining multiple tools.
  • The only tool directly supporting multidimensional metadata exploration (e.g., time, region, media sources).
  • Single-page layout integrates semantic discovery and data visualization seamlessly.
• Experimental or Evaluation Results:
  • Randomized comparative tests show that CorpSum's design significantly enhances user usability (SUS score increased by 43.12 points compared to traditional tools).
  • Provides a comprehensive corpus research tool supporting metadata-driven, multi-level interaction, highly praised by participants.
• Limitations and Future Directions:
  • The current interface only supports English; future plans include multi-language support.
  • Annotation functionality is currently limited to the document level; future versions will explore annotations for words and sentences.
  • More complex visualization designs are not yet included; future evaluations may consider introducing innovative visual components.
  • The sample size was limited; future studies could expand the scale to explore deeper usability issues.

Conclusion and Insights

• Core Contributions:
  • Provides an innovative approach to designing linguistic tools, breaking existing design limitations.
  • Builds a user-oriented, flexible, and efficient linguistic resource analysis platform, advancing academic research tools based on human-computer interaction principles.
  • Proposes 10 design guidelines (e.g., multi-query, intuitive filtering) to inform future tool development.
• Future Work:
  • Enhance interface customization support and expand metadata for more languages.
  • Continue improving usability and refine functionality design based on long-term user feedback.