Design and Application of Computer Aided Translation System Using Self-Attention Based Gated Recurrent Unit

The computer has a strong capability thereby it reduces the labor essential for translation and enhance the effectiveness of transition model. Furthermore, the translators have significant to an intelligent translation software and it tries to apply Artificial Intelligence (AI) for translation. This research proposes a Self-Attention based Gated Recurrent Unit (SA-GRU) for Computer Aided Translation (CAT). The self-attention changes the information importance of input sentence by computing the attention scores which facilitate the model to capture relevant contextual cues. The gating mechanism also serves a purpose of controlling the flow of information to and from the model in the sense that only contained relevant information will remain under preserving the irrelevant information through exclusion. Combining of self-attention using GRU help the model to resolve local and global context in text thus providing higher translation accuracy. The stop words removal, stemming and tokenization are applied as preprocessing. The Term Frequency-Inverse Document Frequency (TF-IDF) is applied for feature extraction. Then, the CAT is performed with the help of SA-GRU. The SA-GRU achieved accuracy of 98.27% which is superior than existing approaches.