Speaker Diarization for Audio Transcription
This research is concerned with speaker diarization for the purpose of facilitating automated speech transcription. This problem has multiple depths depending on the prior knowledge provided to the system. The type and amount of information about the number and characteristics of the speakers can differentiate this problem in a range from a 1-to-N matching, where the voice is compared against different templates, to a clustering problem, where no prior knowledge is available. We intend to find a solution for the speaker diarization problem by incorporating state-of-the-art supervised and unsupervised machine learning methods. This internship will help the interns gain professional work experience in the field of natural language processing with the help from experts in both industry and academia. It is an opportunity for them to practice and improve their industry skills and gain a better understanding of what they are learning in the academia.