![]() | Only 14 pages are availabe for public view |
Abstract Machine translation systems play a significant role in reducing communication barriers between individuals with different languages, especially when human translators are not available or their services are expensive. In spite of the many translation systems that have been developed and translation approaches that have been proposed, machine translation remains a big challenge. This thesis aims to study, develop, and implement a neural machine translation system that achieves a high translation performance. A neural machine translation (NMT) model employs an End-to-End single large neural network, which receives an input sentence and produces its corresponding output sentence. Given a dataset of N sentence pairs, the proposed model is trained to learn a set of parameters 𝜃 that maximize the log-likelihood function. Considering the natural human languages, many problems arise when developing translation systems. First: Human languages are ambiguous; several words have multiple meanings depending on the context in which they occur. Second: The Named Entity problem; some common words in a language are used as proper nouns in another language. |