Skip to main content

Hidden alignment conditional random field, a discriminative string edit distance

Project description

Hidden alignment conditional random field for classifying string pairs - a learnable edit distance.

This package aims to implement the HACRF machine learning model with a sklearn-like interface. It includes ways to fit a model to training examples and score new example.

The model takes string pairs as input and classify them into any number of classes. In McCallum’s original paper the model was applied to the database deduplication problem. Each database entry was paired with every other entry and the model then classified whether the pair was a ‘match’ or a ‘mismatch’ based on training examples of matches and mismatches.

I also tried to use it as learnable string edit distance for normalizing noisy text. See A Conditional Random Field for Discriminatively-trained Finite-state String Edit Distance by McCallum, Bellare, and Pereira, and the report Conditional Random Fields for Noisy text normalisation by Dirko Coetsee.


from pyhacrf import StringPairFeatureExtractor, Hacrf

training_X = [('helloooo', 'hello'), # Matching examples
              ('h0me', 'home'),
              ('krazii', 'crazy'),
              ('non matching string example', 'no really'), # Non-matching examples
              ('and another one', 'yep')]
training_y = ['match',

# Extract features
feature_extractor = StringPairFeatureExtractor(match=True, numeric=True)
training_X_extracted = feature_extractor.fit_transform(training_X)

# Train model
model = Hacrf(l2_regularization=1.0), training_y)

# Evaluate
from sklearn.metrics import confusion_matrix
predictions = model.predict(training_X_extracted)

print(confusion_matrix(training_y, predictions))
> [[0 3]
>  [2 0]]

> [[ 0.94914812  0.05085188]
>  [ 0.92397711  0.07602289]
>  [ 0.86756034  0.13243966]
>  [ 0.05438812  0.94561188]
>  [ 0.02641275  0.97358725]]


This package depends on numpy. The LBFGS optimizer in pylbfgs is used, but alternative optimizers can be passed.


Install by running:

python install

or from pypi:

pip install pyhacrf


Clone from repository, then

pip install -r requirements-dev.txt
cython pyhacrf/*.pyx
python install

To deploy to pypi, make sure you have compiled the *.pyx files to *.c

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pyhacrf-0.1.2.tar.gz (134.1 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page