Gary McGraw & Douglas R. Hofstadter. (1993) Perception and Creation of Alphabetic Style. In Artificial Intelligence and Creativity: Papers from the 1993 Spring Symposium, AAAI Technical Report SS-93-01, AAAI Press.

Author of the summary: Patrawadee Prasangsit, 1999, pp@cc.gatech.edu

Cite this paper for:

• The Letter Spirit project explores the creative act of artistic letter-design.
• Fluid concept is the belief that creativity is an automatic outcome of the existence of sufficiently flexible and context-sensitive concepts.

This paper models how the 26 lowercase letters of the roman alphabet can be rendered in many different but internally coherent styles.

Fluid concept is the belief that creativity is an automatic outcome of the existence of sufficiently flexible and context-sensitive concepts.

The Letter Spirit program contains four global memory structures, each concerned with different levels of concreteness and abstraction of shapes (and concepts pertaining to shapes).  They are:

• Scratchpad: An external memory on which all the letters are drawn and modified.
• Visual Focus: The site where perception of a given letterform occurs. Quanta are fused together into parts, which in turn are interpreted as roles.  Then roles and relations among them suggest membership in one or more letter-categories.  At the beginning, the process is bottom-up.  As structures are built up, top-down influences enter the picture.
• Thematic Focus: The site where the theme of style comes to be established and recognized.  A newly designed letter, though its design is based on a seed letter, has unexpected attributes that are not in the seed letter.  These stylistic attributes are unpredictable emergent by-products of the creation of new letters.  This mean new letterforms give rise to new emergent attributes, which in turn give rise to new letterforms, and so on.
• Conceptual Memory: Provides every concept in the domain with an internal definitional structure (specification in terms of simpler concepts) and a conceptual neighborhood (links to peer concepts in conceptual space).

• Imaginer: Designs in role-level, such as, 'f' has vertical bar, a hook on top right, and a short crossbar, 'x' is two kissing angle-brackets.  It does not know anything about the grid.
• Drafter: Takes grid-independent inputs from the Imaginer and adapts them to the grid.
• Examiner: Takes the specification of a grid-letter in terms of its quanta and determines which of the 26 letter-categories it belongs to, and how strongly and unambiguously so.  Quanta get chunked into parts, which are in turn matched up with conceptual roles.
• Adjudicator: Concerns with stylistic consistency.  A letter must have the same stylistic qualities as the seed letters as well as any already-generated letters.

Here is an example of how these four components work together to design a style.

• Given a seed letter as one in figure 4-left.
• The Examiner identifies the letter by chunking the quanta together to form two roles, a post and a hook.  The roles together suggest two wholes, 'f' and 'l', however they fit 'f' better.
• The seed letter is identified as an 'f'.  A main stylistic guideline "crossbar suppressed" is noted.
• Since 't' is linked as similar to 'f', it is a reasonable one to be designed next.
• The Imaginer applies the current guideline "crossbar suppressed" to designing 't'.
• The Drafter converts the design into a grid-oriented letter, resulting in one like figure 4-middle.
• The Examiner looks at it and determines it as an 'l' (while it's intended as a 't').  The Examiner's diagnosis is that a crossbar is missing.
• The Imaginer fixes this by making creative slippage by having a half-sized crossbar instead of suppressing it.  That the Imaginer minimizes or underdoes the crossbar (as opposed to suppress it) is an example of fluid concept.
• The Drafter takes the above design and draws a grid-oriented 't', which looks like one in figure 4-right.

Summary author's notes:

• none