Automatic speech recognition (ASR) exists for a number of languages including Danish. Spoken language has thus become a viable input source for language technological applications. It has been shown that translators can speak their translations up to six times faster than typing, with similar, or even better, translation quality. Speech input thus also seems an option to enhance speed and quality in human translation production. The PhD project investigates the potential advantages of combining ASR technology with machine translation (MT) in a computer-assisted translation (CAT) environment.
Taking advantage of existing solutions, the PhD candidate will conduct research into the integration of ASR and MT technologies including possible modes for implementation and evaluation. All developments will be based on open-source software. Using the integrated technologies, the PhD candidate will study the situation where the translator reads a source text on a computer screen and speaks out the translation into a microphone (a process called sight translation). The combined ASR and MT system will be evaluated with the existing browser-based CAT tool at CRITT. The PhD candidate will have:
- good understanding of ASR and MT technology (open-source solutions)
- advanced knowledge of English, knowledge of Danish is an advantage
- knowledge of probability theory and statistical modeling
- interest in cognitive modeling of human translation processes
- advanced programming skills
