Metrics#
Here we describe the available metrics in Rubrix:
Text classification: Metrics for text classification
Token classification: Metrics for token classification
Text classification#
- rubrix.metrics.text_classification.metrics.f1(name, query=None)#
Computes the single label f1 metric for a dataset
- Parameters
name (str) β The dataset name.
query (Optional[str]) β An ElasticSearch query with the [query string syntax](https://rubrix.readthedocs.io/en/stable/guides/queries.html)
- Returns
The f1 metric summary
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.text_classification import f1 >>> summary = f1(name="example-dataset") >>> summary.visualize() # will plot a bar chart with results >>> summary.data # returns the raw result data
- rubrix.metrics.text_classification.metrics.f1_multilabel(name, query=None)#
Computes the multi-label label f1 metric for a dataset
- Parameters
name (str) β The dataset name.
query (Optional[str]) β An ElasticSearch query with the [query string syntax](https://rubrix.readthedocs.io/en/stable/guides/queries.html)
- Returns
The f1 metric summary
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.text_classification import f1_multilabel >>> summary = f1_multilabel(name="example-dataset") >>> summary.visualize() # will plot a bar chart with results >>> summary.data # returns the raw result data
Token classification#
- class rubrix.metrics.token_classification.metrics.ComputeFor(value)#
An enumeration.
- rubrix.metrics.token_classification.metrics.entity_capitalness(name, query=None, compute_for=ComputeFor.PREDICTIONS)#
Computes the entity capitalness. The entity capitalness splits the entity mention shape in 4 groups:
UPPER
: All charactes in entity mention are upper case.LOWER
: All charactes in entity mention are lower case.FIRST
: The first character in the mention is upper case.MIDDLE
: First character in the mention is lower case and at least one other character is upper case.- Parameters
name (str) β The dataset name.
query (Optional[str]) β An ElasticSearch query with the query string syntax
compute_for (Union[str, rubrix.metrics.token_classification.metrics.ComputeFor]) β Metric can be computed for annotations or predictions. Accepted values are
Annotations
andPredictions
. Default toPredictions
.
- Returns
The summary entity capitalness distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import entity_capitalness >>> summary = entity_capitalness(name="example-dataset") >>> summary.visualize()
- rubrix.metrics.token_classification.metrics.entity_consistency(name, query=None, compute_for=ComputeFor.PREDICTIONS, mentions=100, threshold=2)#
Computes the consistency for top entity mentions in the dataset.
Entity consistency defines the label variability for a given mention. For example, a mention first identified in the whole dataset as Cardinal, Person and Time is less consistent than a mention Peter identified as Person in the dataset.
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
compute_for (Union[str, rubrix.metrics.token_classification.metrics.ComputeFor]) β Metric can be computed for annotations or predictions. Accepted values are
Annotations
andPredictions
. Default toPredictions
mentions (int) β The number of top mentions to retrieve.
threshold (int) β The entity variability threshold (must be greater or equal to 2).
- Returns
The summary entity capitalness distribution
Examples
>>> from rubrix.metrics.token_classification import entity_consistency >>> summary = entity_consistency(name="example-dataset") >>> summary.visualize()
- rubrix.metrics.token_classification.metrics.entity_density(name, query=None, compute_for=ComputeFor.PREDICTIONS, interval=0.005)#
Computes the entity density distribution. Then entity density is calculated at record level for each mention as
mention_length/tokens_length
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
compute_for (Union[str, rubrix.metrics.token_classification.metrics.ComputeFor]) β Metric can be computed for annotations or predictions. Accepted values are
Annotations
andPredictions
. Default toPredictions
.interval (float) β The interval for histogram. The entity density is defined in the range 0-1.
- Returns
The summary entity density distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import entity_density >>> summary = entity_density(name="example-dataset") >>> summary.visualize()
- rubrix.metrics.token_classification.metrics.entity_labels(name, query=None, compute_for=ComputeFor.PREDICTIONS, labels=50)#
Computes the entity labels distribution
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
compute_for (Union[str, rubrix.metrics.token_classification.metrics.ComputeFor]) β Metric can be computed for annotations or predictions. Accepted values are
Annotations
andPredictions
. Default toPredictions
labels (int) β The number of top entities to retrieve. Lower numbers will be better performants
- Returns
The summary for entity tags distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import entity_labels >>> summary = entity_labels(name="example-dataset", labels=20) >>> summary.visualize() # will plot a bar chart with results >>> summary.data # The top-20 entity tags
- rubrix.metrics.token_classification.metrics.f1(name, query=None)#
Computes F1 metrics for a dataset based on entity-level.
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
- Returns
The F1 metric summary containing precision, recall and the F1 score (averaged and per label).
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import f1 >>> summary = f1(name="example-dataset") >>> summary.visualize() # will plot three bar charts with the results >>> summary.data # returns the raw result data
To display the results as a table:
>>> import pandas as pd >>> pd.DataFrame(summary.data.values(), index=summary.data.keys())
- rubrix.metrics.token_classification.metrics.mention_length(name, query=None, level='token', compute_for=ComputeFor.PREDICTIONS, interval=1)#
Computes mentions length distribution (in number of tokens).
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
level (str) β The mention length level. Accepted values are βtokenβ and βcharβ
compute_for (Union[str, rubrix.metrics.token_classification.metrics.ComputeFor]) β Metric can be computed for annotations or predictions. Accepted values are
Annotations
andPredictions
. Defaults toPredictions
.interval (int) β The bins or bucket for result histogram
- Returns
The summary for mention token distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import mention_length >>> summary = mention_length(name="example-dataset", interval=2) >>> summary.visualize() # will plot a histogram chart with results >>> summary.data # the raw histogram data with bins of size 2
- rubrix.metrics.token_classification.metrics.token_capitalness(name, query=None)#
Computes the token capitalness distribution
UPPER
: All charactes in the token are upper case.LOWER
: All charactes in the token are lower case.FIRST
: The first character in the token is upper case.MIDDLE
: First character in the token is lower case and at least one other character is upper case.- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
- Returns
The summary for token length distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import token_capitalness >>> summary = token_capitalness(name="example-dataset") >>> summary.visualize() # will plot a histogram with results >>> summary.data # The token capitalness distribution
- rubrix.metrics.token_classification.metrics.token_frequency(name, query=None, tokens=1000)#
Computes the token frequency distribution for a numbe of tokens.
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
tokens (int) β The top-k number of tokens to retrieve
- Returns
The summary for token frequency distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import token_frequency >>> summary = token_frequency(name="example-dataset", token=50) >>> summary.visualize() # will plot a histogram with results >>> summary.data # the top-50 tokens frequency
- rubrix.metrics.token_classification.metrics.token_length(name, query=None)#
Computes the token size distribution in terms of number of characters
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
- Returns
The summary for token length distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import token_length >>> summary = token_length(name="example-dataset") >>> summary.visualize() # will plot a histogram with results >>> summary.data # The token length distribution
- rubrix.metrics.token_classification.metrics.tokens_length(name, query=None, interval=1)#
Computes the text length distribution measured in number of tokens.
- Parameters
name (str) β The dataset name.
query (Optional[str]) β
An ElasticSearch query with the query string syntax
interval (int) β The bins or bucket for result histogram
- Returns
The summary for token distribution
- Return type
rubrix.metrics.models.MetricSummary
Examples
>>> from rubrix.metrics.token_classification import tokens_length >>> summary = tokens_length(name="example-dataset", interval=5) >>> summary.visualize() # will plot a histogram with results >>> summary.data # the raw histogram data with bins of size 5