الفهرس 
Title : A deep learning approach for Quran machine translation from Arabic to Italian
AbstractOnly 14 pages are availabe for public view
 |  | from | 99 |  |  |
| | | from | 99 | | |
Abstract
Recently, fast growth in social media platforms makes easier communication between users. According to that, communication increased the importance of translating human languages. Machine Translation (MT) is an important sub-field of Natural Language Processing because it aims to translate natural languages using computers. Machine Translation has become an urgent necessity due to the difference between the languages used within the societies of the world. There are different approaches used to translate natural language, such as rule-based, statistical Machine Translation, and, more recently, Neural Machine Translation. The quality of Machine Translation depends on the availability of parallel datasets. Languages that don’t have enough datasets present many challenges related to their processing and analysis. These languages are defined as low-resource languages. Therefore, many Machine Translation thesis used large parallel corpora to address sets of major European languages. However, only a few research works have considered Italian and Arabic. This research aims to focus on low-resource languages, particularly Arabic language and Italian language. The challenges of low source languages like Arabic, which is a rich morphology language that has different functions of the same words, so that the Arabic content, including the morphological structures of the language, differentiate according to the context of the sentence. While the challenges of the Italian language need a deep study due to the differences in the Italian dialects according to the region as well as the spoken language varies greatly in different parts of Italy. Moreover, dictionary-based translations of the meaning of the Holy Quran from Arabic to Italian are usually incorrect because the dictionary-based translation considers the Quran to be a traditional text and translates it in order of that. This thesis aims to contribute in this regard in two ways. First, it presents a parallel corpus of Italian-Arabic sentences. Second, the thesis introduces two deep learning models, which are long-short-term memory (LSTM) sequence-to-sequence with an attention mechanism and Gated Recurrent Units (GRU) sequence-to-sequence with an attention mechanism for Arabic to Italian Machine Translation. Each of the proposed models is evaluated based on BLEU, ROUGE, and Cosine Similarity scores.
ii
The results indicate that the LSTM-based Neural Machine Translation (NMT) outperforms the GRU-based NMT framework. The experimental results indicate that the LSTM-seq2seq model achieved a score of 0.90 for ROUGE, a score of 0.96 for Cosine Similarity and an average score for BLEU is 0.91. The GRU seq2seq model achieved average scores for BLEU is 0.89 and 0.94, 0.88 for Cosine Similarity, and ROUGE scores, respectively