Paper #15 - Ambiguous Intentions: a Paper-like Interface for Creative Design

Paper:
Ambiguous Intentions: a Paper-like Interface for Creative Design
(Mark D. Gross and Ellen Yi-Luen Do)

Summary:
The authors of the paper create a pen-based interface called Electronic Cocktail Napkin, an interface which handles freehand input by: acquiring its ambiguous and precision information, internally representing it, and visually and edit-behaviorally echoing it to users. Created as a design application, its goal is to support informal drawings by aiming to not only support creation, modification, and management mechanisms of diagrams, but to also use freehand drawings as core information retrieval information. The approach includes mechanisms for: recognizing configurations (i.e., group of elements, which may or may not be drawn abstractly by the user), handling ambiguity and context (i.e., maintaining a representation of unknown and ambiguous elements until it is resolved by context), representing imprecision with constraints (i.e., using an internal, constraint representation which provides interactive edit behavior), and refining and formalizing incrementally (i.e., making improvements with the ultimate design aim of having the program making definite decisions).

Three main mechanisms comprise the Cocktail Napkin application: low level recognition of glyphs, high level recognition of configurations, and constraint and spatial relations maintenance of drawing elements. Low level recognition consists of identifying the best matches given an input’s context and certainty value, and training its recognizer can be done by giving examples to it. Configuration recognition consists of finding patterns in all element pairs of the drawing, and constructing its recognizer is done by extracting constraints from the element types and its spatial relations from the trained examples. Context is maintained as a list of “current context” and “current context chain” in Napkin, where a chain specifies a sequence of other contributing context. Recognition of the previous two mechanisms is done through this context, and recognizing context begins with initial context of only shapes, alphanumeric characters, and lines. Context is then adjusted based on new drawings.

The Napkin application’s interface was tested on undergraduate students and design instructors at the University of Colorado’s architecture program. New testers of the application were given half an hour to familiarize with the application. It was discovered that users primarily wanted to draw, therefore ignoring the recognition facilities. This resulted in focusing on low level tuning and tweaking of the program instead.

Discussion:
An interesting aspect that I found in the Napkin interface was the authors’ aim in creating a system that handles like a very flexible CAD system. With the amount of work done to implement the aim of the authors to employ a smart paper-like system, I’m not sure if it’s enough for the target audience to make a complete switch-over to conventional methods of sketching it into traditional drawing program though. Napkin seems a little too generic with its current implementation, and would need some additions or improvements, such as broadening its functionality as a design program and making use of a sketching language, to be taken more seriously by its target audience.