sklearn.feature_extraction.text.TfidfVectorizer (2024)

Convert a collection of raw documents to a matrix of TF-IDF features.

Equivalent to CountVectorizer followed byTfidfTransformer.

For an example of usage, seeClassification of text documents using sparse features.

For an efficiency comparison of the different feature extractors, seeFeatureHasher and DictVectorizer Comparison.

Read more in the User Guide.

Parameters:

input{‘filename’, ‘file’, ‘content’}, default=’content’

• If 'filename', the sequence passed as an argument to fit isexpected to be a list of filenames that need reading to fetchthe raw content to analyze.

• If 'file', the sequence items must have a ‘read’ method (file-likeobject) that is called to fetch the bytes in memory.

• If 'content', the input is expected to be a sequence of items thatcan be of type string or byte.

encodingstr, default=’utf-8’

If bytes or files are given to analyze, this encoding is used todecode.

decode_error{‘strict’, ‘ignore’, ‘replace’}, default=’strict’

Instruction on what to do if a byte sequence is given to analyze thatcontains characters not of the given encoding. By default, it is‘strict’, meaning that a UnicodeDecodeError will be raised. Othervalues are ‘ignore’ and ‘replace’.

strip_accents{‘ascii’, ‘unicode’} or callable, default=None

Remove accents and perform other character normalizationduring the preprocessing step.‘ascii’ is a fast method that only works on characters that havea direct ASCII mapping.‘unicode’ is a slightly slower method that works on any characters.None (default) means no character normalization is performed.

Both ‘ascii’ and ‘unicode’ use NFKD normalization fromunicodedata.normalize.

lowercasebool, default=True

Convert all characters to lowercase before tokenizing.

preprocessorcallable, default=None

Override the preprocessing (string transformation) stage whilepreserving the tokenizing and n-grams generation steps.Only applies if analyzer is not callable.

tokenizercallable, default=None

Override the string tokenization step while preserving thepreprocessing and n-grams generation steps.Only applies if analyzer == 'word'.

analyzer{‘word’, ‘char’, ‘char_wb’} or callable, default=’word’

Whether the feature should be made of word or character n-grams.Option ‘char_wb’ creates character n-grams only from text insideword boundaries; n-grams at the edges of words are padded with space.

If a callable is passed it is used to extract the sequence of featuresout of the raw, unprocessed input.

Changed in version 0.21: Since v0.21, if input is 'filename' or 'file', the datais first read from the file and then passed to the given callableanalyzer.

stop_words{‘english’}, list, default=None

If a string, it is passed to _check_stop_list and the appropriate stoplist is returned. ‘english’ is currently the only supported stringvalue.There are several known issues with ‘english’ and you shouldconsider an alternative (see Using stop words).

If a list, that list is assumed to contain stop words, all of whichwill be removed from the resulting tokens.Only applies if analyzer == 'word'.

If None, no stop words will be used. In this case, setting max_dfto a higher value, such as in the range (0.7, 1.0), can automatically detectand filter stop words based on intra corpus document frequency of terms.

token_patternstr, default=r”(?u)\b\w\w+\b”

Regular expression denoting what constitutes a “token”, only usedif analyzer == 'word'. The default regexp selects tokens of 2or more alphanumeric characters (punctuation is completely ignoredand always treated as a token separator).

If there is a capturing group in token_pattern then thecaptured group content, not the entire match, becomes the token.At most one capturing group is permitted.

ngram_rangetuple (min_n, max_n), default=(1, 1)

The lower and upper boundary of the range of n-values for differentn-grams to be extracted. All values of n such that min_n <= n <= max_nwill be used. For example an ngram_range of (1, 1) means onlyunigrams, (1, 2) means unigrams and bigrams, and (2, 2) meansonly bigrams.Only applies if analyzer is not callable.

max_dffloat or int, default=1.0

When building the vocabulary ignore terms that have a documentfrequency strictly higher than the given threshold (corpus-specificstop words).If float in range [0.0, 1.0], the parameter represents a proportion ofdocuments, integer absolute counts.This parameter is ignored if vocabulary is not None.

min_dffloat or int, default=1

When building the vocabulary ignore terms that have a documentfrequency strictly lower than the given threshold. This value is alsocalled cut-off in the literature.If float in range of [0.0, 1.0], the parameter represents a proportionof documents, integer absolute counts.This parameter is ignored if vocabulary is not None.

max_featuresint, default=None

If not None, build a vocabulary that only consider the topmax_features ordered by term frequency across the corpus.Otherwise, all features are used.

This parameter is ignored if vocabulary is not None.

vocabularyMapping or iterable, default=None

Either a Mapping (e.g., a dict) where keys are terms and values areindices in the feature matrix, or an iterable over terms. If notgiven, a vocabulary is determined from the input documents.

binarybool, default=False

If True, all non-zero term counts are set to 1. This does not meanoutputs will have only 0/1 values, only that the tf term in tf-idfis binary. (Set binary to True, use_idf to False andnorm to None to get 0/1 outputs).

dtypedtype, default=float64

Type of the matrix returned by fit_transform() or transform().

norm{‘l1’, ‘l2’} or None, default=’l2’

Each output row will have unit norm, either:

• ‘l2’: Sum of squares of vector elements is 1. The cosinesimilarity between two vectors is their dot product when l2 norm hasbeen applied.

• ‘l1’: Sum of absolute values of vector elements is 1.See normalize.

• None: No normalization.

use_idfbool, default=True

Enable inverse-document-frequency reweighting. If False, idf(t) = 1.

smooth_idfbool, default=True

Smooth idf weights by adding one to document frequencies, as if anextra document was seen containing every term in the collectionexactly once. Prevents zero divisions.

sublinear_tfbool, default=False

Apply sublinear tf scaling, i.e. replace tf with 1 + log(tf).

Attributes:

vocabulary_dict

A mapping of terms to feature indices.

See Also

TF-IDF — Term Frequency-Inverse Document Frequency – LearnDataSciTF-IDF: Weighing Importance in Text - Let's Data ScienceTF-IDF: An Introduction | Built InAn Introduction to TF-IDF: What It Is & How to Use It

fixed_vocabulary_bool

True if a fixed vocabulary of term to indices mappingis provided by the user.

idf_array of shape (n_features,)

Inverse document frequency vector, only defined if use_idf=True.

stop_words_set

Terms that were ignored because they either:

• occurred in too many documents (max_df)

• occurred in too few documents (min_df)

• were cut off by feature selection (max_features).

This is only available if no vocabulary was given.

Notes

The stop_words_ attribute can get large and increase the model sizewhen pickling. This attribute is provided only for introspection and canbe safely removed using delattr or set to None before pickling.

Examples

>>> from sklearn.feature_extraction.text import TfidfVectorizer>>> corpus = [...  'This is the first document.',...  'This document is the second document.',...  'And this is the third one.',...  'Is this the first document?',... ]>>> vectorizer = TfidfVectorizer()>>> X = vectorizer.fit_transform(corpus)>>> vectorizer.get_feature_names_out()array(['and', 'document', 'first', 'is', 'one', 'second', 'the', 'third', 'this'], ...)>>> print(X.shape)(4, 9)

Methods

build_analyzer()[source]¶

Return a callable to process input data.

The callable handles preprocessing, tokenization, and n-grams generation.

Returns:

analyzer: callable

A function to handle preprocessing, tokenizationand n-grams generation.

build_preprocessor()[source]¶

Return a function to preprocess the text before tokenization.

Returns:

preprocessor: callable

A function to preprocess the text before tokenization.

build_tokenizer()[source]¶

Return a function that splits a string into a sequence of tokens.

Returns:

tokenizer: callable

A function to split a string into a sequence of tokens.

decode(doc)[source]¶

Decode the input into a string of unicode symbols.

The decoding strategy depends on the vectorizer parameters.

Parameters:

docbytes or str

The string to decode.

Returns:

doc: str

A string of unicode symbols.

fit(raw_documents, y=None)[source]¶

Learn vocabulary and idf from training set.

Parameters:

raw_documentsiterable

An iterable which generates either str, unicode or file objects.

yNone

This parameter is not needed to compute tfidf.

Returns:

selfobject

Fitted vectorizer.

fit_transform(raw_documents, y=None)[source]¶

Learn vocabulary and idf, return document-term matrix.

This is equivalent to fit followed by transform, but more efficientlyimplemented.

Parameters:

raw_documentsiterable

An iterable which generates either str, unicode or file objects.

yNone

This parameter is ignored.

Returns:

Xsparse matrix of (n_samples, n_features)

Tf-idf-weighted document-term matrix.

get_feature_names_out(input_features=None)[source]¶

Get output feature names for transformation.

See Also

TF-IDF Explained And Python Sklearn Implementation

Parameters:

input_featuresarray-like of str or None, default=None

Not used, present here for API consistency by convention.

Returns:

feature_names_outndarray of str objects

Transformed feature names.

get_metadata_routing()[source]¶

Get metadata routing of this object.

Please check User Guide on how the routingmechanism works.

Returns:

routingMetadataRequest

A MetadataRequest encapsulatingrouting information.

get_params(deep=True)[source]¶

Get parameters for this estimator.

Parameters:

deepbool, default=True

If True, will return the parameters for this estimator andcontained subobjects that are estimators.

Returns:

paramsdict

Parameter names mapped to their values.

get_stop_words()[source]¶

Build or fetch the effective stop words list.

Returns:

stop_words: list or None

A list of stop words.

property idf_¶

Inverse document frequency vector, only defined if use_idf=True.

Returns:

ndarray of shape (n_features,)

inverse_transform(X)[source]¶

Return terms per document with nonzero entries in X.

Parameters:

X{array-like, sparse matrix} of shape (n_samples, n_features)

Document-term matrix.

Returns:

X_invlist of arrays of shape (n_samples,)

List of arrays of terms.

set_fit_request(*, raw_documents: bool | None | str = '$UNCHANGED$') → TfidfVectorizer[source]¶

Request metadata passed to the fit method.

Note that this method is only relevant ifenable_metadata_routing=True (see sklearn.set_config).Please see User Guide on how the routingmechanism works.

The options for each parameter are:

• True: metadata is requested, and passed to fit if provided. The request is ignored if metadata is not provided.

• False: metadata is not requested and the meta-estimator will not pass it to fit.

• None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.

• str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains theexisting request. This allows you to change the request for someparameters and not others.

New in version 1.3.

Note

This method is only relevant if this estimator is used as asub-estimator of a meta-estimator, e.g. used inside aPipeline. Otherwise it has no effect.

Parameters:

raw_documentsstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED

Metadata routing for raw_documents parameter in fit.

Returns:

selfobject

The updated object.

set_params(**params)[source]¶

Set the parameters of this estimator.

The method works on simple estimators as well as on nested objects(such as Pipeline). The latter haveparameters of the form <component>__<parameter> so that it’spossible to update each component of a nested object.

Parameters:

**paramsdict

Estimator parameters.

Returns:

selfestimator instance

Estimator instance.

set_transform_request(*, raw_documents: bool | None | str = '$UNCHANGED$') → TfidfVectorizer[source]¶

Request metadata passed to the transform method.

Note that this method is only relevant ifenable_metadata_routing=True (see sklearn.set_config).Please see User Guide on how the routingmechanism works.

The options for each parameter are:

• True: metadata is requested, and passed to transform if provided. The request is ignored if metadata is not provided.

• False: metadata is not requested and the meta-estimator will not pass it to transform.

• None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.

• str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains theexisting request. This allows you to change the request for someparameters and not others.

New in version 1.3.

Note

This method is only relevant if this estimator is used as asub-estimator of a meta-estimator, e.g. used inside aPipeline. Otherwise it has no effect.

Parameters:

raw_documentsstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED

Metadata routing for raw_documents parameter in transform.

Returns:

selfobject

The updated object.

transform(raw_documents)[source]¶

Transform documents to document-term matrix.

Uses the vocabulary and document frequencies (df) learned by fit (orfit_transform).

Parameters:

raw_documentsiterable

An iterable which generates either str, unicode or file objects.

Returns:

Xsparse matrix of (n_samples, n_features)

Tf-idf-weighted document-term matrix.