On-line recognition of Pitman's handwritten shorthand—an evaluation of potential

Abstract

This paper describes a number of evaluation experiments designed to establish the potential of Pitman's handwritten shorthand as an input for computer transcription to text. Such a system would have applications in verbatim reporting, the electronic office, and as an aid for the deaf. The experiments described compare the performance of a proposed computer transcription system for shorthand (described previously in this journal) with the benchmark performance obtained using human transcription. In addition, measurements on typical Pitman shorthand data are used to estimate potential performance limits. It is concluded that the poor overall performance of the proposed computer transcription system is due to a combination of three factors: first, the simplified nature of the recognition algorithms used compared with the knowledge-based techniques used by human shorthand readers; secondly, ergonomie deficiencies of the data input devices used; and finally, writer's lack of familiarity with the system and its capabilities. A proposed strategy for improving the performance of the system by attention to the first two of these deficiencies is given.