huggingface/rasa_hmtl

RASA wrapper for HMTL: Hierarchical Multi-Task Learning

RASA wrapper for HMTL (Hierarchical Multi-Task Learning)

🌊 A State-of-the-Art neural network model for several NLP tasks based on PyTorch and AllenNLP

---

@article{sanh2018hmtl,
  title={A Hierarchical Multi-task Approach for Learning Embeddings from Semantic Tasks},
  author={Sanh, Victor and Wolf, Thomas and Ruder, Sebastian},
  journal={arXiv preprint arXiv:1811.06031},
  year={2018}
}

⚠ Work in progress, this has not been thoroughly tested. ⚠

Main repo: https://github.com/huggingface/hmtl Demo: https://huggingface.co/hmtl/

This code sample demonstrates how to use rasa_nlu's Component mechanism to integrate the tasks results from HMTL:

• Named Entity Recoginition
• Entity Mention Detection
• Relation Extraction
• Coreference Resolution

rasa_hmtl's Issues

ModuleNotFoundError: No module named 'overrides'

Hi guys

Thank you for continually producing and sharing these fantastic components!

I'm trying to incorporate HMTL in the Rasa NLU pipeline.

I downloaded and installed AllenNLP in a dedicated virtual environment, installed the latest Rasa 1.0, followed by installation of HMTL.

I modified the scripts/machine_setup.sh to install HMTL (minus the python 3.6 and the virtual env, since python 3.6.7 is already installed and accessible from the current virtual env). I also commented out boto* in the requirements.txt since I don't intend to interact w/ AWS.

I also modified a few lines because some of the AllenNLP modules have either been moved or removed in the latest release. Right now, the following statement in rasa_hmtl.py:

from hmtl.hmtlPredictor import HMTLPredictor

errors out with msg:

ModuleNotFoundError: No module named 'hmtl'

Since this a user supplied module, I assume I should manually specify the path, but it's not obvious to me what I need to add.

Thank you for your advice.

Best,
H