Word Prediction for Amharic Online Handwriting Recognition
No Thumbnail Available
Date
2008-07
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Online handwriting recognition, keypads, soft keys are some of the data entry techniques used
to enter data into mobile devices, such as smart phone, PDA etc. Data entry in these devices
could be either predictive or non-predictive. A word prediction method is a data entry
technique in which the first few characters of the word is written and the remaining are
predicted. Among the data entry techniques, online handwriting recognition is commonly
used for handheld devices such as PDAs. When online handwriting recognition is combined
with word prediction, the data entry process will be more efficient.
In this work, we have proposed a word prediction model for Amharic online handwriting
recognition. To design the model: a corpus of 131,399 Amharic words is prepared to extract
statistical information that is used to determine the value of N for the N-gram model, where
the value two (2) is considered as a result of the analyses made a combination of an Amharic
dictionary (lexicon) and a list of names of persons and places with a total size of 17,137 has
been used.
To show the validity of the word prediction model and the algorithm designed, a prototype is
developed. Experiment is also conducted to measure the accuracy of the word prediction
engine and a prediction accuracy of 81.39% is achieved.
Keywords: Amharic Online Handwriting Recognition, Amharic word prediction model,
N-gram model for Amharic word prediction, word prediction corpus
Description
Keywords
Amharic Online Handwriting Recognition; Amharic Word Prediction Model; N-Gram Model for Amharic Word Prediction; Word Prediction Corpus