An English to Urdu translation model based on CBR, ANN and translation rules

Abstract

In this paper, a model has been presented for English to Urdu machine translation based on case-based reasoning CBR technique for machine translation MT, translation rule base model and artificial neural network ANN model. This paper describes the architecture and working of the MT system. We have used translation rules-based case marking for inflection of words according to the number, person and gender in the target language. The CBR approach is used as learning technique for the selection of Urdu translation rules for the input English sentence. The integration of rule-based model with CBR combines the representation and reasoning. Neural network adds the retrieval and learning processes in the computation for machine translation. The rule-based model enhances the adaptation process. The translation results obtained from the system have been evaluated and achieved an average n-gram BLEU score 0.728, meteor score 0.869, F-measure score 0.894, unigram precision 0.923 and unigram recall 0.909.