TL;DR
This paper introduces Inky, a novel command-line interface for web tasks that combines the efficiency of typed commands with rich visual feedback, aiming to mitigate the usability issues of traditional command lines. By allowing "sloppy" syntax and providing real-time interpretations and contextual information, Inky offers a hybrid approach between command-line efficiency and GUI usability for web automation.
What Problem Does This Paper Address?
Traditional web interfaces often rely on menu-and-form structures, which are great for learnability but can be inefficient for frequent users, especially with poorly designed or lengthy forms. Command-line interfaces (CLIs) offer efficiency but come with a steep learning curve, requiring users to memorize commands and syntax, leading to potential errors. This paper addresses the gap between these two paradigms by proposing a system that captures the speed of CLIs while minimizing their complexity and error proneness, specifically for web automation tasks. The authors also highlight the challenge of "information scraps" – data that users jot down outside of structured PIM tools due to the high cost of data entry, and aim to provide a lightweight capture mechanism.
Methods
Inky employs a "sloppy" command-line approach, meaning it has little to no strict syntax to learn. Commands can have keywords in any order, command names can be omitted or replaced with synonyms, and parameters can be entered in various ways. The core of Inky is a keyword command interpreter that uses a dynamic programming algorithm to match user input against a database of known functions and types.
Key novel techniques include:
- Rich Visual Feedback: As the user types, Inky continuously displays a ranked list of possible interpretations of the command. This feedback includes icons for the target website, arguments already provided, and clearly marked missing parameters.
- Sloppy Syntax Handling: Inky uses type and function recognizers to interpret commands, allowing for synonyms and flexible ordering of keywords and arguments. A post-processing step handles unquoted strings.
- Hybrid Interface: Inky integrates GUI elements into its feedback, such as date pickers or lists of conference rooms, to aid parameter entry and provide context.
- Two Modes of Execution: Commands can be run in "view mode" (default, fills the form but doesn't submit) or "submit mode" (Control-Enter, executes the side-effect immediately). This distinction helps prevent accidental side effects.
- Self-Disclosure: Inky provides reminders of its capabilities by showing example commands on startup and suggesting Inky commands when users interact with equivalent web functions.
The paper also introduces "Pinky," a successor prototype focused on PIM data capture, which further integrates GUI widgets and web clippings for enhanced context and ease of use.
Datasets
Not discussed in the paper. The system is evaluated based on user studies and performance measurements of the implemented prototypes.
Key Insights and Ideas
- The Power of "Sloppy" Syntax: Users benefit greatly from not having to adhere to rigid syntax. Inky's ability to handle synonyms and flexible ordering significantly improves accuracy and reduces the cognitive load.
- Rich Feedback is Crucial: Real-time visual feedback on command interpretation, including missing parameters, dramatically aids learnability and reduces errors. It acts as a form of "just-in-time" learning.
- Bridging CLI and GUI: The hybrid approach, where a CLI leverages GUI elements for input and context, offers a compelling path forward for efficient web interaction. This is particularly relevant for tasks where quick data entry is paramount.
- Contextual Information Reduces Errors: Displaying relevant information (like calendar availability or recent emails) directly within the command interface helps users make better decisions and avoid mistakes, especially for tasks with side effects.
- Self-Disclosure as a Usability Feature: Reminding users of Inky's capabilities when they are already interacting with the underlying web functionality can subtly encourage adoption and learning.
Key Results
- Learnability: Inky was found to be fairly learnable with minimal instructions. Four out of seven users reported learning commands through startup suggestions.
- Accuracy: The Inky prototype achieved an overall accuracy rate of 73% for correctly interpreted commands. With synonym support, this accuracy increased to 81%. Without synonyms, it dropped to 61%.
- Synonym Importance: The study strongly validated the importance of synonym support, a feature often overlooked in practice.
- Order Independence: While users naturally tended to place function keywords at the start, Inky's support for order independence was deemed important.
- Suggestion Usability: Users found multiple similar suggestions confusing. They often preferred to adjust their command to match the top suggestion rather than scan alternatives. This suggests a need for fewer, more distinct suggestions.
- Performance: The keyword engine achieved response times of around 15 milliseconds on average, with a maximum of 55 milliseconds, which is more than adequate for real-time feedback. Performance scaled well with command length and database size.
- User Adoption: Only two out of seven users ran commands in "submit mode" (which executes side-effects automatically), indicating a cautious approach to side-effect commands, which is supported by the default "view mode."
Limitations and Future Work
- Suggestion Clarity: The study indicated that alternative interpretations often looked too similar, making them hard to compare. Future work should focus on reducing the number of suggestions and differentiating them more clearly.
- User Extension: The current system requires editing XML files to add new functions. Future work could involve connecting Inky to web macro recording systems (like Smart Bookmarks or Koala/CoScripter) to allow end-users to extend Inky with new web site functions more easily.
- Pinky Development: The Pinky prototype, with its enhanced GUI widgets and web clippings, is still under development and requires further testing.
- Open-Source Release: The authors plan to release Inky and Pinky as open-source software.
The paper concludes by emphasizing the desirability of building command lines on top of existing graphical user interfaces to provide fast data entry and reduced navigation, a reversal of historical UI trends.
Summary updated