TY  - RPRT
TI  - Truly unsupervised acoustic word embeddings using weak top-down constraints in encoder-decoder models
AU  - Herman Kamper
PY  - 2019
UR  - https://arxiv.org/abs/1811.00403
ID  - 1811.00403
ER  -