Post-processing of Handwritten Pitman's Shorthand Using Unigram and Heuristic Approaches

Abstract

The computer transcription of handwritten Pitman's shorthand has enormous potential as a means of rapid text entry to today's handheld devices. Recognition errors caused in pattern segmentation and classification raises the incidence of ambiguous interpretation in existing systems and the paper proposes a well-established unigram technique and an efficient heuristic method to reduce ambiguity in a linguistic post processor. Heuristics applied in our transcription system are: - firstly, incorporating visual stimulus as used by human readers; secondly, applying knowledge of the most common words of Pitman shorthand; and finally, adding knowledge of collocation. An experiment using a phonetic Lexicon of 5000 entries shows the distribution of ambiguity in a shorthand lexicon due to the similarity of outlines' and estimates the transcription accuracy of 94%.