We perform text normalization, i.e. the transformation of words from the written to the spoken form, using a memory augmented neural network. With the addition of dynamic memory access and storage mechanism, we present a neural architecture that will serve as a language-agnostic text normalization system while avoiding the kind of unacceptable errors made by the LSTM-based recurrent neural networks. By successfully reducing the frequency of such mistakes, we show that this novel architecture is indeed a better alternative. Our proposed system requires significantly lesser amounts of data, training time and compute resources. Additionally, we perform data up-sampling, circumventing the data sparsity problem in some semiotic classes, to show that sufficient examples in any particular class can improve the performance of our text normalization system. Although a few occurrences of these errors still remain in certain semiotic classes, we demonstrate that memory augmented networks with meta-learning capabilities can open many doors to a superior text normalization system.

No abstract

No abstract

Sending encrypted messages frequently will draw the attention of third parties, i.e. crackers andhackers, perhaps causing attempts to break and reveal the original messages. In a digital world, steganography is introduced to hide the existence of the communication by concealing a secret message inside another unsuspicious message.The aim of this paper isto produce proposed method to provide a high level security system by implementing and designing multi-level steganography system to hide data in a color video-cover. This system is more complex system, it implemented in two levels of embedding and this is an issue of the high level of security because it required two levels of extraction to extract the hidden data. The system is implemented in the frequency domain, using wavelet transform domain. The idea of using transformation in the proposed system is due to the results of previous published works which indicated the hiding in the frequency domain is more effective than hiding in time domain, due to compactness attributes of some transforms and due to its robustness. A singular value decomposition (SVD) is also used in this proposed system. MATLAB programming environment is used to simulate the total system.

No abstract