The aim of this thesis is to investigate and compare tools for speech-to-text conversion and foreign language translation in the teaching process. A tool that performs this function correctly enough could be used for people studying abroad, who have not yet fully mastered the language of the lecturers.
The first part of the thesis describes the development of speech recognition and conversion tools from the oldest systems to the latest technologies, such as Alexa and Siri. This is followed by a presentation of the development of tools for machine translation into foreign languages – from the initial translations using dictionaries to the most modern translation method using a neural network.
The following section explains how speech-to-text conversion and translation tools work. We look at all three tools discussed in the thesis, i.e., Google Docs with voice typing, Google Translate with conversation mode and Online Notes. The latter was developed by a team of researchers led by the Faculty of Computer and Information Science in collaboration with the Centre for Language Resources and Technologies of the University of Ljubljana.
After the presentation of the three tools, we see the success of their performance. We are given the specifications of the evaluated recordings followed by the results of the usage tests.
In the analytical part, observations from the tests and practical examples of speech recognition and translation errors are presented. It is evident that the tools perform much worse in the translation of science-based than social-based lectures.
Finally, a comparison of the usability of the three tools is made. We find that the two most commercial tools are the least suitable for pedagogical use.
|
