Paper #24 - Properties of Real-World Digital Logic Diagrams

Paper:
Properties of Real-World Digital Logic Diagrams
(Christine Alvarado and Michael Lazzareschi)

Summary:
The authors of this paper strive to tackle the problem of creating a sketch recognition system that handles the trade-off between ease of recognition and drawing freedom. The focus is on how students naturally draw digital logic diagrams for an educational design of a recognition system. Three aspects are stroke order, stroke timing, and stroke number. The motivation stems from students drawing differently from experienced designers, and also from little work done on the digital logic domain that presents an important challenge for recognizers. The authors state their contributions in discovering that drawing patterns of students violate standard restrictions up to 34%, having analysis that informs free-sketch sketch recognition systems for digital circuit diagrams, and also having analysis that provides a framework to investigate the belief that similar drawing patterns exist in other domains.

In the data collection and analysis portion of the paper, a complete set of sketches and notes were taken from digital design and computer architecture students using tablet computers. Stroke order was done by counting how many symbols students drew and didn’t draw that were temporarily contiguous. Stroke timing was done by comparing the pause time between consecutive strokes in the same symbol. Stroke number was done by computing the number of strokes drawn for each symbol.

Different results were produced from the three aspects above. For stroke order, two different drawing patterns were seen. Students either drew a majority of a symbol with consecutive strokes or came back to that symbol later to add a single stroke due to correction or overtracing, or the student left the symbol obviously unfinished for later completion to invert the output. For stroke time, it turned out that students paused longer when switching symbols. Pauses varied per student, and error bars of pauses showed that it was impossible to find a reliable time threshold to indicate a new symbol being drawn. For stroke number, the number of strokes varied greatly per student. Students varied in consistency of drawing each individual type of symbol and rarely used a single stroke to span multiple symbols. The authors note that such trends may be domain-specific. In conclusion, the authors believe that pause times can aid stroke grouping, user-specific learning may aid recognition, recognition systems need to identify symbols that are not drawn with consecutive strokes, and symbol recognizers must incorporate a wide range of drawing styles.

Discussion:
Out of all the papers we've read thus far in this course, I found it most interesting that the sketch recognition domain in this paper may be the most specific of the bunch. Previous papers we've read have also analyzed sketch recognition using digital logic as a domain (Oltmanns, Sezgin, etc.), but this paper deals more specifically on drawing styles by students. Some people may say that the contributions from this paper would be more limited since the domain is narrower, but I believe that this paper may serve as a useful reference in developing sketch recognizers targeting university students in order to enhance current curriculum methods.