Evaluating Google Speech-to-Text API's Performance for Romanian e-Learning Resources

Abstract

This paper presents a way of performing ASR on multimedia e-learning resources available in Romanian with the usage of the Google Cloud Speech-to-Text API. The material presents the history of ASR systems together with the main approaches used by the algorithms behind these systems. The cloud computing providers, that offer ASR solutions via SaaS, are analyzed as well. After performing a short literature review, the author focuses on applying the Google Cloud Speech-to-Text API on various video e-learning resources available online on YouTube. By doing this, the resources can be easily indexed and transformed into searchable materials. The WER score is used in order to measure the accuracy of the model and to compare it with similar works. The results are more than satisfying, thus the proposed model can be used as a method of automating the indexing of multimedia e-learning resources.