Abstract

This research paper conducted a deep dive into the field of computational linguistics with a focus on machine translation, or translation conducted by a computer. I explored how machine translation tools like Google Translate function, the various benefits and ethical concerns that come with these tools, and how to make these tools more effective for users.

My research showed that machine translation in its present state has several ethical issues, and while I believe that these tools are beneficial, problems like bias, safety, and quality should be fixed to remove any ethical concerns. Machine translation typically makes use of neural machine translation, which relies on large amounts of data to “think” like a human and form sentences. As such, the data given can be biased to a particular gender (gender bias) or have other biases that can affect machine translation. In addition, many modern linguistic-related developments rely on large-language models (LLMs), which not only power some machine translation tools like Google Translate but also popular AI tools such as ChatGPT. LLMs perform better and have varying amounts of safety depending on the language used, which are both part of linguistic discrimination. Despite the problems, some of these issues mentioned can be solved by adding additional processes like adding “tags” to the data an LLM uses, and machine translation has been influential in allowing real-time translations of less-used languages, leading to a change in an increasingly homogenous world.

My work was significant in furthering the discussion of ethics in machine translation. Some papers focus on simply the quality issues that come with machine translation, but neglect to focus on other issues like the ones mentioned earlier. Based on my research and findings, developers of machine translation tools and LLMs should place a greater focus on safety and ethics to mitigate ethical concerns in machine translation.