The first census optical character recognition system conference

Survey of the State of the Art in Human Language Technology

“…In one of our recent studies, we discovered that a good typist tolerates only up to error using a special keyboard that introduced random typing errors at a softwarecontrollable rate; 0.5% error is unnoticeable; ¦ error is intolerable! (Warwick, 1995) Human subjects make 4-8% error for isolated letters read in the absence of context and R ¢ ¡ error with the context of the neighboring letters (Wilkinson, Geist, et al, 1992;Geist et al, 1994). Therefore, the task of designing usable handwriting recognizers for pen computing applications is tremendously hard.…”

An unconstrained handwriting recognition system

“…In the first one the system was trained with symbols from the subsets HSF 0-7 of the NIST [37] database and tested on the same database. In the second case the system trained with the NIST but was tested on the IAM-DB [26].…”

Effectiveness of feature and classifier algorithms in character recognition systems

“…At the first Census OCR System Conference a large number of systems (40 for digits) were used to recognize the same sample of characters [1]. Neural network systems, systems combining neural network methods with other methods (hybrid system), and systems based entirely on statistical pattern recognition methods were used.…”