seqeval

seqeval is a Python framework for sequence labeling evaluation. seqeval can evaluate the performance of chunking tasks such as named-entity recognition, part-of-speech tagging, semantic role labeling and so on.

This is well-tested by using the Perl script conlleval, which can be used for measuring the performance of a system that has processed the CoNLL-2000 shared task data.

Support features

seqeval supports following schemes:

• IOB1
• IOB2
• IOE1
• IOE2
• IOBES(only in strict mode)
• BILOU(only in strict mode)

and following metrics:

| metrics | description | |---|---| | accuracy_score(y_true, y_pred) | Compute the accuracy. | | precision_score(y_true, y_pred) | Compute the precision. | | recall_score(y_true, y_pred) | Compute the recall. | | f1_score(y_true, y_pred) | Compute the F1 score, also known as balanced F-score or F-measure. | | classification_report(y_true, y_pred, digits=2) | Build a text report showing the main classification metrics. digits is number of digits for formatting output floating point values. Default value is 2. |

Usage

seqeval supports the two evaluation modes. You can specify the following mode to each metrics:

• default
• strict

The default mode is compatible with conlleval. If you want to use the default mode, you don't need to specify it:

>>> from seqeval.metrics import accuracy_score
>>> from seqeval.metrics import classification_report
>>> from seqeval.metrics import f1_score
>>> y_true = [['O', 'O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]
>>> y_pred = [['O', 'O', 'B-MISC', 'I-MISC', 'I-MISC', 'I-MISC', 'O'], ['B-PER', 'I-PER', 'O']]
>>> f1_score(y_true, y_pred)
0.50
>>> classification_report(y_true, y_pred)
              precision    recall  f1-score   support

        MISC       0.00      0.00      0.00         1
         PER       1.00      1.00      1.00         1

   micro avg       0.50      0.50      0.50         2
   macro avg       0.50      0.50      0.50         2
weighted avg       0.50      0.50      0.50         2

In strict mode, the inputs are evaluated according to the specified schema. The behavior of the strict mode is different from the default one which is designed to simulate conlleval. If you want to use the strict mode, please specify mode='strict' and scheme arguments at the same time:

>>> from seqeval.scheme import IOB2
>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)
              precision    recall  f1-score   support

        MISC       0.00      0.00      0.00         1
         PER       1.00      1.00      1.00         1

   micro avg       0.50      0.50      0.50         2
   macro avg       0.50      0.50      0.50         2
weighted avg       0.50      0.50      0.50         2

A minimum case to explain differences between the default and strict mode:

>>> from seqeval.metrics import classification_report
>>> from seqeval.scheme import IOB2
>>> y_true = [['B-NP', 'I-NP', 'O']]
>>> y_pred = [['I-NP', 'I-NP', 'O']]
>>> classification_report(y_true, y_pred)
              precision    recall  f1-score   support
          NP       1.00      1.00      1.00         1
   micro avg       1.00      1.00      1.00         1
   macro avg       1.00      1.00      1.00         1
weighted avg       1.00      1.00      1.00         1
>>> classification_report(y_true, y_pred, mode='strict', scheme=IOB2)
              precision    recall  f1-score   support
          NP       0.00      0.00      0.00         1
   micro avg       0.00      0.00      0.00         1
   macro avg       0.00      0.00      0.00         1
weighted avg       0.00      0.00      0.00         1

Installation

To install seqeval, simply run:

pip install seqeval

License

MIT

Citation

@misc{seqeval,
  title={{seqeval}: A Python framework for sequence labeling evaluation},
  url={https://github.com/chakki-works/seqeval},
  note={Software available from https://github.com/chakki-works/seqeval},
  author={Hiroki Nakayama},
  year={2018},
}