Susanto et al. (2016)#

Publication#

Learning to Capitalize with Character-Level Recurrent Neural Networks: An Empirical Study

Repositories#

The original repository for the paper is this one, but our implementation wraps a PyTorch version here that is based loosely on the original paper.

Available Models#

The repository provides truecasing models for four different languages:

  • English

    • Description: An English model trained on Wikipedia

    • Name: susanto2016-truecaser

    • Usage:

      from repro.models.susanto2016 import RNNTruecaser
model = RNNTruecaser("wiki-truecaser-model-en.tar.gz")
truecased = model.predict("text")

  • Spanish

    • Description: An Spanish model trained on WMT

    • Name: susanto2016-truecaser

    • Usage:

      from repro.models.susanto2016 import RNNTruecaser
model = RNNTruecaser("wmt-truecaser-model-es.tar.gz")
truecased = model.predict("text")

  • German

    • Description: A German model trained on WMT

    • Name: susanto2016-truecaser

    • Usage:

      from repro.models.susanto2016 import RNNTruecaser
model = RNNTruecaser("wmt-truecaser-model-de.tar.gz")
truecased = model.predict("text")

  • Russian

    • Description: A Russian model trained on LORELEI data

    • Name: susanto2016-truecaser

    • Usage:

      from repro.models.susanto2016 import RNNTruecaser
model = RNNTruecaser("lrl-truecaser-model-ru.tar.gz")
truecased = model.predict("text")

Implementation Notes#

Docker Information#

  • Image name: susanto2016

  • Build command:

    repro setup susanto2016 [--silent]

  • Requires network: No

Testing#

repro setup susanto2016
pytest models/susanto2016/tests

Status#

  • [x] Regression unit tests pass
    See here

  • [ ] Correctness unit tests pass
    No expected output provided

  • [ ] Model runs on full test dataset
    Not tested

  • [ ] Predictions approximately replicate results reported in the paper
    Not tested

  • [ ] Predictions exactly replicate results reported in the paper
    Not tested