Contextual Urdu Lemmatization Using Recurrent Neural Network Models

In the field of natural language processing, machine translation is a colossally developing research area that helps humans communicate more effectively by bridging linguistic gap. translation, normalization and morphological analyses are first perhaps most important modules for information retrieval (IR). To build analyzer, or to complete process, it extract correct root out different words. Stemming lemmatization techniques commonly used find words in language. However, few studies on IR systems Urdu have shown effective than stemming due infixes found This paper presents algorithm based recurrent neural network models resource-scarce languages such as not very common. The proposed model trained tested two datasets, namely, Monolingual Corpus (UMC) Universal Dependencies (UDU). datasets lemmatized with help models. Word2Vec edit trees generate semantic syntactic embedding. Bidirectional long short-term memory (BiLSTM), bidirectional gated unit (BiGRU), (BiGRNN), attention-free encoder–decoder (AFED) under defined hyperparameters. Experimental results show encoder-decoder achieves an accuracy, precision, recall, F-score 0.96, 0.95, respectively, outperforms existing