tatk.nlu.svm.camrest package

Submodules

tatk.nlu.svm.camrest.evaluate module

Evaluate SVMNLU models on Camrest test dataset

Metric:

dataset level Precision/Recall/F1

Usage:

PYTHONPATH=../../../.. python evaluate.py [usr|sys|all]

tatk.nlu.svm.camrest.evaluate.da2triples(dialog_act)

tatk.nlu.svm.camrest.nlu module

SVMNLU build a classifier for each semantic tuple (intent-slot-value) based on n-gram features. It’s first proposed by Mairesse et al. (2009). We adapt the implementation from pydial. For more information, please refer to tatk/nlu/svm/camrest/README.md

Trained models can be download on:

Reference:

Mairesse, F., Gasic, M., Jurcicek, F., Keizer, S., Thomson, B., Yu, K., & Young, S. (2009, April). Spoken language understanding from unaligned data using discriminative classification models. In 2009 IEEE International Conference on Acoustics, Speech and Signal Processing (pp. 4749-4752). IEEE.

class tatk.nlu.svm.camrest.nlu.SVMNLU(mode)

Bases: tatk.nlu.nlu.NLU

__init__(mode)

SVM NLU initialization.

Args:
mode (str):

can be either ‘usr’, ‘sys’ or ‘all’, representing which side of data the model was trained on.

Example:

nlu = SVMNLU(mode=’usr’)

predict(utterance, context=[])

Predict the dialog act of a natural language utterance.

Args:
utterance (str):

A natural language utterance.

Returns:
output (dict):

The dialog act of utterance.

tatk.nlu.svm.camrest.preprocess module

Preprocess camrest data for SVMNLU.

Usage:

python preprocess [mode=all|usr|sys] mode: which side data will be use

Require:

  • ../../../../data/camrest/[train|val|test].json.zip data file

  • ../../../../data/camrest/db database dir

Output:

  • configs/ontology_camrest_[mode].json ontology file

  • data/[mode]_data/ processed data dir

tatk.nlu.svm.camrest.preprocess.read_zipped_json(filepath, filename)