WORDGRAPH: Keyword-in-Context Visualization for NETSPEAK's Wildcard Search.

Area of Science:

• Natural Language Processing
• Human-Computer Interaction
• Information Retrieval

Background:

• Writers often struggle with selecting optimal phrases during text composition.
• Existing tools may lack efficient mechanisms for exploring phrase alternatives and their contextual relevance.

Purpose of the Study:

• To introduce and evaluate WORDGRAPH, a novel system for visual phrase selection and retrieval.
• To assess the effectiveness of WORDGRAPH's graphical interface compared to a textual interface for various retrieval tasks.

Main Methods:

• Implementation of a scalable retrieval engine supporting probabilistic strategies for high-quality, millisecond response times.
• Development of a graphical user interface (WORDGRAPH) alongside a traditional textual list display.
• Conducting a user study to compare interface performance across phrase verification, context-sensitive word choice, and corpus exploration tasks.

Main Results:

• The textual interface is adequate for simple phrase verification.
• Both interfaces support context-sensitive word choice effectively.
• WORDGRAPH excels in exploring phrase context and the underlying corpus, especially for queries with multiple wildcards.

Conclusions:

• WORDGRAPH offers a superior interactive experience for exploring linguistic data and enhancing writer's choices.
• The system's visual approach and efficient retrieval engine benefit writers, particularly in complex phrase-finding scenarios.