Handwriting detector, extractor, and language classifier
摘要
說(shuō)明書(shū)
In the prediction module 1920, information 1990 for a particular handwriting sample may be input to the feature determination module 1950. Feature determination module 1950b may determine the same set of features (e.g. frequency metrics) or a different set of features from the information 1990 as feature determination module 1950a determined from handwriting data store 1740. In some examples, feature determination module 1950a and 1950b are the same module. Feature determination module 1950b produces feature vector 1915, which is input into the model 1980 to generate a probability indicator that the information 1990 is of a particular language type. In some aspects, the model 1980 generates multiple probabilities, each probability indicating a probability that the information 1990 is of a particular language type. The training module 1910 may operate in an offline manner to train the model 1980. The prediction module 1920, however, may be designed to operate in an online manner. It should be noted that the model 1980 may be periodically updated via additional training and/or user feedback.
