Large vocabulary recognition for online Turkish handwriting with sublexical units

Abstract

We present a system for large vocabulary recognition of online Turkish handwriting, using hidden Markovmodels. While using a traditional approach for the recognizer, we have identified and developed solutions for the mainproblems specific to Turkish handwriting recognition. First, since large amounts of Turkish handwriting samples are notavailable, the system is trained and optimized using the large UNIPEN dataset of English handwriting, before extendingit to Turkish using a small Turkish dataset. The delayed strokes, which pose a significant source of variation in writingorder due to the large number of diacritical marks in Turkish, are removed during preprocessing. Finally, as a solution tothe high out-of-vocabulary rates encountered when using a fixed size lexicon in general purpose recognition, a lexicon isconstructed from sublexical units (stems and endings) learned from a large Turkish corpus. A statistical bigram languagemodel learned from the same corpus is also applied during the decoding process.The system obtains a 91.7% word recognition rate when tested on a small Turkish handwritten word datasetusing a medium sized (1950 words) lexicon corresponding to the vocabulary of the test set and 63.8% using a large,general purpose lexicon (130,000 words). However, with the proposed stem+ending lexicon (12,500 words) and bigramlanguage model with lattice expansion, a 67.9% word recognition accuracy is obtained, surpassing the results obtainedwith the general purpose lexicon while using a much smaller one