Probe

Discriminative feature-based information retrieval and an NLP feature engineering framework

Current version: 0.9.2

If you are using it in an academic work, please cite

@inproceedings{chen2017discriminative,
  title={Discriminative information retrieval for question answering sentence selection},
  author={Chen, Tongfei and Van Durme, Benjamin},
  booktitle={Proceedings of EACL},
  year={2017}
}

Probe contains two modules:

• probe-core: compositional, declarative framework for feature engineering
• probe-search: Implements feature-based discriminative IR based on Apache Lucene and probe-core.

Guide to each can be found in their respective readme files.

Changelog

• 0.9.2: Added Featurizer.>>>(FeaturizerFamily).
• 0.9.1: Added Featurizer.tagWith.
• 0.9.0: Supports weighted indices through Lucene's payload mechanism; update Lucene to 6.5.0.
• 0.8.0: Merged probe-search from a standalone package.
• 0.7.4: Added Featurizer.optional.
• 0.7.3: Maintenance version bump for depending libraries.
• 0.7.2: Bugfix: escape space to underscore in feature vector formatters.
• 0.7.1: Various dot operations on feature families.
• 0.7.0: Feature families.