2022 01 26

• Should we upload the JSTPS paper on arXiv?
• The previous idea of using the CLIP model to train an audio encoder has been already explored in a couple of recent papers:
  • Wu, Ho-Hsiang, et al. "Wav2CLIP: Learning Robust Audio Representations From CLIP." arXiv preprint arXiv:2110.11499 (2021). link
  • Zhao, Yanpeng, et al. "Connecting the Dots between Audio and Text without Parallel Data through Visual Knowledge Transfer." arXiv preprint arXiv:2112.08995 (2021). link
• Discuss new direction for future work:
  • Apply our framework for language documentation (the Yorùbá language)