ABSTRACT
Artificial intelligence (AI) is a big field of studies that enables computers and machines to mimic the perception, learning, problem-solving, and decision-making capabilities of humans. More recently, with increasing computing power and data, AI has pushed the boundaries of human-like intelligence. In this chapter, we will discuss how AI can be used to understand text contents to perform translation.
The main challenge in AI-based translation comes from the difficulties in understanding the semantics of texts as well as understanding the syntactic structure of texts to translate them. We start from the traditional rule-based machine translation approach to show how human-crafted rules can be used to translate texts and illustrate their limitations. We then move on to statistical machine translation, which learns to translate automatically from text corpus without too much human intervention. After that, with the development of the big data and machine learning technologies, we further discuss the recent deep learning–based approaches in translation and show their strengths. The key component of deep learning is neural networks, which are used to extract text semantics and learn translation patterns from the training data. Importantly, we discuss the widely used attention mechanism, which enables a very deep understanding of the word-level text and sentence relationships to enable precise translation. Finally, we elaborate on the recent advances in language and translation models in the industry and discuss their potential future directions and the current limitations of machine-based translation techniques.
