"""Embedding Coherence Test metric implementation."""
from typing import Any, Callable, Dict, List, Union
import numpy as np
from scipy.spatial.distance import cosine
from scipy.stats import spearmanr
from wefe.metrics.base_metric import BaseMetric
from wefe.preprocessing import get_embeddings_from_query
from wefe.query import Query
from wefe.word_embedding_model import WordEmbeddingModel
[docs]class ECT(BaseMetric):
"""Embedding Coherence Test (ECT).
The metric was originally proposed in [1] and implemented in [2].
It calculates the average target group vectors, measures the cosine similarity of
each to a list of attribute words and calculates the correlation of the resulting
similarity lists.
Values closer to 1 are better as they represent less bias.
The general steps of the test, as defined in [1], are as follows:
1. Embed all given target and attribute words with the given embedding model.
2. Calculate mean vectors for the two sets of target word vectors.
3. Measure the cosine similarity of the mean target vectors to all of the given
attribute words.
4. Calculate the Spearman r correlation between the resulting two lists of
similarities.
5. Return the correlation value as score of the metric (in the range of -1 to 1);
higher is better.
Values closer to 1 are better as they represent less bias.
References
----------
| [1]: Dev, S., & Phillips, J. (2019, April). Attenuating Bias in Word vectors.
| [2]: https://github.com/sunipa/Attenuating-Bias-in-Word-Vec
"""
# The metrics accepts two target sets and a single attribute set
metric_template = (2, 1)
metric_name = "Embedding Coherence Test"
metric_short_name = "ECT"
[docs] def run_query(
self,
query: Query,
model: WordEmbeddingModel,
lost_vocabulary_threshold: float = 0.2,
preprocessors: List[Dict[str, Union[str, bool, Callable]]] = [{}],
strategy: str = "first",
normalize: bool = False,
warn_not_found_words: bool = False,
*args: Any,
**kwargs: Any
) -> Dict[str, Any]:
"""Run ECT with the given query with the given parameters.
Parameters
----------
query : Query
A Query object that contains the target and attribute word sets to be
tested.
model : WordEmbeddingModel
A word embedding model.
preprocessors : List[Dict[str, Union[str, bool, Callable]]]
A list with preprocessor options.
A ``preprocessor`` is a dictionary that specifies what processing(s) are
performed on each word before it is looked up in the model vocabulary.
For example, the ``preprocessor``
``{'lowecase': True, 'strip_accents': True}`` allows you to lowercase
and remove the accent from each word before searching for them in the
model vocabulary. Note that an empty dictionary ``{}`` indicates that no
preprocessing is done.
The possible options for a preprocessor are:
* ``lowercase``: ``bool``. Indicates that the words are transformed to
lowercase.
* ``uppercase``: ``bool``. Indicates that the words are transformed to
uppercase.
* ``titlecase``: ``bool``. Indicates that the words are transformed to
titlecase.
* ``strip_accents``: ``bool``, ``{'ascii', 'unicode'}``: Specifies that
the accents of the words are eliminated. The stripping type can be
specified. True uses 'unicode' by default.
* ``preprocessor``: ``Callable``. It receives a function that operates
on each word. In the case of specifying a function, it overrides the
default preprocessor (i.e., the previous options stop working).
A list of preprocessor options allows you to search for several
variants of the words into the model. For example, the preprocessors
``[{}, {"lowercase": True, "strip_accents": True}]``
``{}`` allows searching first for the original words in the vocabulary of
the model. In case some of them are not found,
``{"lowercase": True, "strip_accents": True}`` is executed on these words
and then they are searched in the model vocabulary.
strategy : str, optional
The strategy indicates how it will use the preprocessed words: 'first' will
include only the first transformed word found. 'all' will include all
transformed words found, by default "first".
normalize : bool, optional
True indicates that embeddings will be normalized, by default False
warn_not_found_words : bool, optional
Specifies if the function will warn (in the logger)
the words that were not found in the model's vocabulary, by default False.
Returns
-------
Dict[str, Any]
A dictionary with the query name and the result of the query.
Examples
--------
>>> from wefe.metrics import ECT
>>> from wefe.query import Query
>>> from wefe.utils import load_test_model
>>>
>>> # define the query
>>> query = Query(
... target_sets=[
... ["female", "woman", "girl", "sister", "she", "her", "hers",
... "daughter"],
... ["male", "man", "boy", "brother", "he", "him", "his", "son"],
... ],
... attribute_sets=[
... [
... "home", "parents", "children", "family", "cousins", "marriage",
... "wedding", "relatives",
... ],
... ],
... target_sets_names=["Female terms", "Male Terms"],
... attribute_sets_names=["Family"],
... )
>>> # load the model (in this case, the test model included in wefe)
>>> model = load_test_model()
>>>
>>> # instance the metric and run the query
>>> ECT().run_query(query, model) # doctest: +SKIP
{'query_name': 'Female terms and Male Terms wrt Family',
'result': 0.6190476190476191,
'ect': 0.6190476190476191}
>>> # if you want the embeddings to be normalized before calculating the metrics
>>> # use the normalize parameter as True before executing the query.
>>> ECT().run_query(query, model, normalize=True) # doctest: +SKIP
{'query_name': 'Female terms and Male Terms wrt Family',
'result': 0.7619047619047621,
'ect': 0.7619047619047621}
"""
# check the types of the provided arguments (only the defaults).
self._check_input(query, model, locals())
# transform query word sets into embeddings
embeddings = get_embeddings_from_query(
model=model,
query=query,
lost_vocabulary_threshold=lost_vocabulary_threshold,
preprocessors=preprocessors,
strategy=strategy,
normalize=normalize,
warn_not_found_words=warn_not_found_words,
)
# If the lost vocabulary threshold is exceeded, return the default value
if embeddings is None:
return {"query_name": query.query_name, "result": np.nan, "ect": np.nan}
# get the targets and attribute sets transformed into embeddings.
target_sets, attribute_sets = embeddings
# get only the embeddings of the sets.
target_embeddings = list(target_sets.values())
attribute_embeddings = list(attribute_sets.values())
ect = self.__calculate_embedding_coherence(
list(target_embeddings[0].values()),
list(target_embeddings[1].values()),
list(attribute_embeddings[0].values()),
)
return {"query_name": query.query_name, "result": ect, "ect": ect}
def __calculate_embedding_coherence(
self, target_set_1: list, target_set_2: list, attribute_set: list
) -> float:
"""Calculate the ECT metric over the given parameters. Return the result.
Parameters
----------
target_set_1 : list
The first set of target words.
target_set_2 : list
The second set of target words.
attribute_set : list
The set of attribute words.
Returns
-------
float
The value denoting the Spearman correlation.
"""
# Calculate mean vectors for both target vector sets
target_means = [np.mean(s, axis=0) for s in (target_set_1, target_set_2)]
# Measure similarities between mean vecotrs and all attribute words
similarities = []
for mean_vector in target_means:
similarities.append([1 - cosine(mean_vector, a) for a in attribute_set])
# Calculate similarity correlations
return spearmanr(similarities[0], similarities[1]).correlation