Training and evaluating denoising autoencoder implementations to generate noise-free sentence embeddings from sentences with simulated word substitution errors.
Training script fond in src/run-all-training.py
Note: this requires a package that is yet to be published (asr-error-simulator). Contact me if interested in setting up and running the code.
.
├── AUTHORS.md
├── LICENSE
├── README.md
├── bin
├── config
├── data
│ ├── external
│ ├── interim
│ ├── processed
│ └── raw
├── docs
├── notebooks
├── reports
│ └── figures
└── src
├── data
├── external
├── models
├── tools
└── visualization