The field of interactive prediction of neural machine translation (IPNMT) is defined as a mode of machine translation which allows real-time interaction between the translator and the machine translation system. Often, human translators need to accomplish real-time translations quickly. They also might face issues finding the right words, encounter repetitive words that require translation, and face other challenges such as character limits when typing translations from a language with shorter average sentence lengths to a longer one. The IPNMT can help human translators translate from one language to another more efficiently through word suggestions, suggest the top few relevant words to reduce cognitive load, and provide a diversity of suggestions. Additionally, along with human translators, language learners would also find machine-guided translation useful when interacting with a tool that can provide on-the-fly suggestions. This can help language learners expand their vocabulary and assist them in understanding the language.
The goal of this project is to explore various approaches to create an interactive prediction tool that can suggest next words during a translation using Neural Machine Translation (NMT) based models and compare the translation quality against static machine translation to evaluate how adding the interactivity can affect the translation quality. Our task is limited only to the French-English translation system and generating the translations from left-to-right, where the user can accept or correct the suggested prefix from the system. The input would be the French sentence (source) and the resulting translation would be in English (target). The model generates new words based on the corrected prefix until the sentence translation is completed. Finally, the goal of this project also involves exploring state-of-the art NMT models, understanding the structure of French and English languages and how they affect the outputs from the model through data-exploration, and retrieving insights from the experiments that were run.