gensim logo

gensim
gensim tagline

Get Expert Help From The Gensim Authors

Consulting in Machine Learning & NLP

• Commercial document similarity engine: ScaleText.ai

Corporate trainings in Python Data Science and Deep Learning

utils – Various utility functions

utils – Various utility functions

Various general utility functions.

class gensim.utils.ClippedCorpus(corpus, max_docs=None)

Bases: gensim.utils.SaveLoad

Wrap a corpus and return max_doc element from it.

Parameters:
  • corpus (iterable of iterable of (int, numeric)) – Input corpus.
  • max_docs (int) – Maximum number of documents in the wrapped corpus.

Warning

Any documents after max_docs are ignored. This effectively limits the length of the returned corpus to <= max_docs. Set max_docs=None for “no limit”, effectively wrapping the entire input corpus.

classmethod load(fname, mmap=None)

Load an object previously saved using save() from a file.

Parameters:
  • fname (str) – Path to file that contains needed object.
  • mmap (str, optional) – Memory-map option. If the object was saved with large arrays stored separately, you can load these arrays via mmap (shared memory) using mmap=’r’. If the file being loaded is compressed (either ‘.gz’ or ‘.bz2’), then `mmap=None must be set.

See also

save()
Save object to file.
Returns:Object loaded from fname.
Return type:object
Raises:AttributeError – When called on an object instance instead of class (this is a class method).
save(fname_or_handle, separately=None, sep_limit=10485760, ignore=frozenset([]), pickle_protocol=2)

Save the object to a file.

Parameters:
  • fname_or_handle (str or file-like) – Path to output file or already opened file-like object. If the object is a file handle, no special array handling will be performed, all attributes will be saved to the same file.
  • separately (list of str or None, optional) –

    If None, automatically detect large numpy/scipy.sparse arrays in the object being stored, and store them into separate files. This prevent memory errors for large objects, and also allows memory-mapping the large arrays for efficient loading and sharing the large arrays in RAM between multiple processes.

    If list of str: store these attributes into separate files. The automated size check is not performed in this case.

  • sep_limit (int, optional) – Don’t store arrays smaller than this separately. In bytes.
  • ignore (frozenset of str, optional) – Attributes that shouldn’t be stored at all.
  • pickle_protocol (int, optional) – Protocol number for pickle.

See also

load()
Load object from file.
class gensim.utils.FakeDict(num_terms)

Bases: object

Objects of this class act as dictionaries that map integer->str(integer), for a specified range of integers <0, num_terms).

This is meant to avoid allocating real dictionaries when num_terms is huge, which is a waste of memory.

Parameters:num_terms (int) – Number of terms.
get(val, default=None)
iteritems()

Iterate over all keys and values.

Yields:(int, str) – Pair of (id, token).
keys()

Override the dict.keys(), which is used to determine the maximum internal id of a corpus, i.e. the vocabulary dimensionality.

Returns:Highest id, packed in list.
Return type:list of int

Notes

To avoid materializing the whole range(0, self.num_terms), this returns the highest id = [self.num_terms - 1] only.

class gensim.utils.InputQueue(q, corpus, chunksize, maxsize, as_numpy)

Bases: multiprocessing.process.Process

Populate a queue of input chunks from a streamed corpus.

Useful for reading and chunking corpora in the background, in a separate process, so that workers that use the queue are not starved for input chunks.

Parameters:
  • q (multiprocessing.Queue) – Enqueue chunks into this queue.
  • corpus (iterable of iterable of (int, numeric)) – Corpus to read and split into “chunksize”-ed groups
  • chunksize (int) – Split corpus into chunks of this size.
  • as_numpy (bool, optional) – Enqueue chunks as numpy.ndarray instead of lists.
authkey
daemon

Return whether process is a daemon

exitcode

Return exit code of process or None if it has yet to stop

ident

Return identifier (PID) of process or None if it has yet to start

is_alive()

Return whether process is alive

join(timeout=None)

Wait until child process terminates

name
pid

Return identifier (PID) of process or None if it has yet to start

run()

Method to be run in sub-process; can be overridden in sub-class

start()

Start child process

terminate()

Terminate process; sends SIGTERM signal or uses TerminateProcess()

class gensim.utils.RepeatCorpus(corpus, reps)

Bases: gensim.utils.SaveLoad

Wrap a corpus as another corpus of length reps. This is achieved by repeating documents from corpus over and over again, until the requested length len(result) == reps is reached. Repetition is done on-the-fly=efficiently, via itertools.

Examples

>>> from gensim.utils import RepeatCorpus
>>>
>>> corpus = [[(1, 2)], []] # 2 documents
>>> list(RepeatCorpus(corpus, 5)) # repeat 2.5 times to get 5 documents
[[(1, 2)], [], [(1, 2)], [], [(1, 2)]]
Parameters:
  • corpus (iterable of iterable of (int, numeric)) – Input corpus.
  • reps (int) – Number of repeats for documents from corpus.
classmethod load(fname, mmap=None)

Load an object previously saved using save() from a file.

Parameters:
  • fname (str) – Path to file that contains needed object.
  • mmap (str, optional) – Memory-map option. If the object was saved with large arrays stored separately, you can load these arrays via mmap (shared memory) using mmap=’r’. If the file being loaded is compressed (either ‘.gz’ or ‘.bz2’), then `mmap=None must be set.

See also

save()
Save object to file.
Returns:Object loaded from fname.
Return type:object
Raises:AttributeError – When called on an object instance instead of class (this is a class method).
save(fname_or_handle, separately=None, sep_limit=10485760, ignore=frozenset([]), pickle_protocol=2)

Save the object to a file.

Parameters:
  • fname_or_handle (str or file-like) – Path to output file or already opened file-like object. If the object is a file handle, no special array handling will be performed, all attributes will be saved to the same file.
  • separately (list of str or None, optional) –

    If None, automatically detect large numpy/scipy.sparse arrays in the object being stored, and store them into separate files. This prevent memory errors for large objects, and also allows memory-mapping the large arrays for efficient loading and sharing the large arrays in RAM between multiple processes.

    If list of str: store these attributes into separate files. The automated size check is not performed in this case.

  • sep_limit (int, optional) – Don’t store arrays smaller than this separately. In bytes.
  • ignore (frozenset of str, optional) – Attributes that shouldn’t be stored at all.
  • pickle_protocol (int, optional) – Protocol number for pickle.

See also

load()
Load object from file.
class gensim.utils.RepeatCorpusNTimes(corpus, n)

Bases: gensim.utils.SaveLoad

Wrap a corpus and repeat it n times.

Examples

>>> from gensim.utils import RepeatCorpusNTimes
>>>
>>> corpus = [[(1, 0.5)], []]
>>> list(RepeatCorpusNTimes(corpus, 3)) # repeat 3 times
[[(1, 0.5)], [], [(1, 0.5)], [], [(1, 0.5)], []]
Parameters:
  • corpus (iterable of iterable of (int, numeric)) – Input corpus.
  • n (int) – Number of repeats for corpus.
classmethod load(fname, mmap=None)

Load an object previously saved using save() from a file.

Parameters:
  • fname (str) – Path to file that contains needed object.
  • mmap (str, optional) – Memory-map option. If the object was saved with large arrays stored separately, you can load these arrays via mmap (shared memory) using mmap=’r’. If the file being loaded is compressed (either ‘.gz’ or ‘.bz2’), then `mmap=None must be set.

See also

save()
Save object to file.
Returns:Object loaded from fname.
Return type:object
Raises:AttributeError – When called on an object instance instead of class (this is a class method).
save(fname_or_handle, separately=None, sep_limit=10485760, ignore=frozenset([]), pickle_protocol=2)

Save the object to a file.

Parameters:
  • fname_or_handle (str or file-like) – Path to output file or already opened file-like object. If the object is a file handle, no special array handling will be performed, all attributes will be saved to the same file.
  • separately (list of str or None, optional) –

    If None, automatically detect large numpy/scipy.sparse arrays in the object being stored, and store them into separate files. This prevent memory errors for large objects, and also allows memory-mapping the large arrays for efficient loading and sharing the large arrays in RAM between multiple processes.

    If list of str: store these attributes into separate files. The automated size check is not performed in this case.

  • sep_limit (int, optional) – Don’t store arrays smaller than this separately. In bytes.
  • ignore (frozenset of str, optional) – Attributes that shouldn’t be stored at all.
  • pickle_protocol (int, optional) – Protocol number for pickle.

See also

load()
Load object from file.
class gensim.utils.SaveLoad

Bases: object

Serialize/deserialize object from disk, by equipping objects with the save()/load() methods.

Warning

This uses pickle internally (among other techniques), so objects must not contain unpicklable attributes such as lambda functions etc.

classmethod load(fname, mmap=None)

Load an object previously saved using save() from a file.

Parameters:
  • fname (str) – Path to file that contains needed object.
  • mmap (str, optional) – Memory-map option. If the object was saved with large arrays stored separately, you can load these arrays via mmap (shared memory) using mmap=’r’. If the file being loaded is compressed (either ‘.gz’ or ‘.bz2’), then `mmap=None must be set.

See also

save()
Save object to file.
Returns:Object loaded from fname.
Return type:object
Raises:AttributeError – When called on an object instance instead of class (this is a class method).
save(fname_or_handle, separately=None, sep_limit=10485760, ignore=frozenset([]), pickle_protocol=2)

Save the object to a file.

Parameters:
  • fname_or_handle (str or file-like) – Path to output file or already opened file-like object. If the object is a file handle, no special array handling will be performed, all attributes will be saved to the same file.
  • separately (list of str or None, optional) –

    If None, automatically detect large numpy/scipy.sparse arrays in the object being stored, and store them into separate files. This prevent memory errors for large objects, and also allows memory-mapping the large arrays for efficient loading and sharing the large arrays in RAM between multiple processes.

    If list of str: store these attributes into separate files. The automated size check is not performed in this case.

  • sep_limit (int, optional) – Don’t store arrays smaller than this separately. In bytes.
  • ignore (frozenset of str, optional) – Attributes that shouldn’t be stored at all.
  • pickle_protocol (int, optional) – Protocol number for pickle.

See also

load()
Load object from file.
class gensim.utils.SlicedCorpus(corpus, slice_)

Bases: gensim.utils.SaveLoad

Wrap corpus and return a slice of it.

Parameters:
  • corpus (iterable of iterable of (int, numeric)) – Input corpus.
  • slice (slice or iterable) – Slice for corpus.

Notes

Negative slicing can only be used if the corpus is indexable, otherwise, the corpus will be iterated over. Slice can also be a np.ndarray to support fancy indexing.

Calculating the size of a SlicedCorpus is expensive when using a slice as the corpus has to be iterated over once. Using a list or np.ndarray does not have this drawback, but consumes more memory.

classmethod load(fname, mmap=None)

Load an object previously saved using save() from a file.

Parameters:
  • fname (str) – Path to file that contains needed object.
  • mmap (str, optional) – Memory-map option. If the object was saved with large arrays stored separately, you can load these arrays via mmap (shared memory) using mmap=’r’. If the file being loaded is compressed (either ‘.gz’ or ‘.bz2’), then `mmap=None must be set.

See also

save()
Save object to file.
Returns:Object loaded from fname.
Return type:object
Raises:AttributeError – When called on an object instance instead of class (this is a class method).
save(fname_or_handle, separately=None, sep_limit=10485760, ignore=frozenset([]), pickle_protocol=2)

Save the object to a file.

Parameters:
  • fname_or_handle (str or file-like) – Path to output file or already opened file-like object. If the object is a file handle, no special array handling will be performed, all attributes will be saved to the same file.
  • separately (list of str or None, optional) –

    If None, automatically detect large numpy/scipy.sparse arrays in the object being stored, and store them into separate files. This prevent memory errors for large objects, and also allows memory-mapping the large arrays for efficient loading and sharing the large arrays in RAM between multiple processes.

    If list of str: store these attributes into separate files. The automated size check is not performed in this case.

  • sep_limit (int, optional) – Don’t store arrays smaller than this separately. In bytes.
  • ignore (frozenset of str, optional) – Attributes that shouldn’t be stored at all.
  • pickle_protocol (int, optional) – Protocol number for pickle.

See also

load()
Load object from file.
gensim.utils.any2unicode(text, encoding='utf8', errors='strict')

Convert text (bytestring in given encoding or unicode) to unicode.

Parameters:
  • text (str) – Input text.
  • errors (str, optional) – Error handling behaviour if text is a bytestring.
  • encoding (str, optional) – Encoding of text if it is a bytestring.
Returns:

Unicode version of text.

Return type:

str

gensim.utils.any2utf8(text, errors='strict', encoding='utf8')

Convert a unicode or bytes string in the given encoding into a utf8 bytestring.

Parameters:
  • text (str) – Input text.
  • errors (str, optional) – Error handling behaviour if text is a bytestring.
  • encoding (str, optional) – Encoding of text if it is a bytestring.
Returns:

Bytestring in utf8.

Return type:

str

gensim.utils.call_on_class_only(*args, **kwargs)

Helper to raise AttributeError if a class method is called on an instance. Used internally.

Parameters:
  • *args – Variable length argument list.
  • **kwargs – Arbitrary keyword arguments.
Raises:

AttributeError – If a class method is called on an instance.

gensim.utils.check_output(stdout=-1, *popenargs, **kwargs)

Run OS command with the given arguments and return its output as a byte string.

Backported from Python 2.7 with a few minor modifications. Widely used for gensim.models.wrappers. Behaves very similar to https://docs.python.org/2/library/subprocess.html#subprocess.check_output.

Examples

>>> from gensim.utils import check_output
>>> check_output(args=['echo', '1'])
'1\n'
Raises:KeyboardInterrupt – If Ctrl+C pressed.
gensim.utils.chunkize(corpus, chunksize, maxsize=0, as_numpy=False)

Split corpus into fixed-sized chunks, using chunkize_serial().

Parameters:
  • corpus (iterable of object) – An iterable.
  • chunksize (int) – Split corpus into chunks of this size.
  • maxsize (int, optional) – If > 0, prepare chunks in a background process, filling a chunk queue of size at most maxsize.
  • as_numpy (bool, optional) – Yield chunks as np.ndarray instead of lists?
Yields:

list OR np.ndarray – “chunksize”-ed chunks of elements from corpus.

Notes

Each chunk is of length chunksize, except the last one which may be smaller. A once-only input stream (corpus from a generator) is ok, chunking is done efficiently via itertools.

If maxsize > 0, don’t wait idly in between successive chunk yields, but rather keep filling a short queue (of size at most maxsize) with forthcoming chunks in advance. This is realized by starting a separate process, and is meant to reduce I/O delays, which can be significant when corpus comes from a slow medium like HDD, database or network.

If maxsize == 0, don’t fool around with parallelism and simply yield the chunksize via chunkize_serial() (no I/O optimizations).

Yields:list of object OR np.ndarray – Groups based on iterable
gensim.utils.chunkize_serial(iterable, chunksize, as_numpy=False, dtype=<type 'numpy.float32'>)

Yield elements from iterable in “chunksize”-ed groups.

The last returned element may be smaller if the length of collection is not divisible by chunksize.

Parameters:
  • iterable (iterable of object) – An iterable.
  • chunksize (int) – Split iterable into chunks of this size.
  • as_numpy (bool, optional) – Yield chunks as np.ndarray instead of lists.
Yields:

list OR np.ndarray – “chunksize”-ed chunks of elements from iterable.

Examples

>>> print(list(grouper(range(10), 3)))
[[0, 1, 2], [3, 4, 5], [6, 7, 8], [9]]

Recursively copy a directory ala shutils.copytree, but hardlink files instead of copying.

Parameters:
  • source (str) – Path to source directory
  • dest (str) – Path to destination directory

Warning

Available on UNIX systems only.

gensim.utils.deaccent(text)

Remove letter accents from the given string.

Parameters:text (str) – Input string.
Returns:Unicode string without accents.
Return type:str

Examples

>>> from gensim.utils import deaccent
>>> deaccent("Šéf chomutovských komunistů dostal poštou bílý prášek")
u'Sef chomutovskych komunistu dostal postou bily prasek'
gensim.utils.decode_htmlentities(text)

Decode all HTML entities in text that are encoded as hex, decimal or named entities. Adapted from python-twitter-ircbot/html_decode.py.

Parameters:text (str) – Input HTML.

Examples

>>> from gensim.utils import decode_htmlentities
>>>
>>> u = u'E tu vivrai nel terrore - L&#x27;aldil&#xE0; (1981)'
>>> print(decode_htmlentities(u).encode('UTF-8'))
E tu vivrai nel terrore - L'aldilà (1981)
>>> print(decode_htmlentities("l&#39;eau"))
l'eau
>>> print(decode_htmlentities("foo &lt; bar"))
foo < bar
gensim.utils.deprecated(reason)

Decorator to mark functions as deprecated.

Calling a decorated function will result in a warning being emitted, using warnings.warn. Adapted from https://stackoverflow.com/a/40301488/8001386.

Parameters:reason (str) – Reason of deprecation.
Returns:Decorated function
Return type:function
gensim.utils.dict_from_corpus(corpus)

Scan corpus for all word ids that appear in it, then construct a mapping which maps each word_id -> str(word_id).

Parameters:corpus (iterable of iterable of (int, numeric)) – Collection of texts in BoW format.
Returns:id2word – “Fake” mapping which maps each word_id -> str(word_id).
Return type:FakeDict

Warning

This function is used whenever words need to be displayed (as opposed to just their ids) but no word_id -> word mapping was provided. The resulting mapping only covers words actually used in the corpus, up to the highest word_id found.

gensim.utils.effective_n_jobs(n_jobs)

Determines the number of jobs can run in parallel.

Just like in sklearn, passing n_jobs=-1 means using all available CPU cores.

Parameters:n_jobs (int) – Number of workers requested by caller.
Returns:Number of effective jobs.
Return type:int
gensim.utils.file_or_filename(input)

Open a filename for reading with smart_open, or seek to the beginning if input is an already open file.

Parameters:input (str or file-like) – Filename or file-like object.
Returns:An open file, positioned at the beginning.
Return type:file-like object
gensim.utils.flatten(nested_list)

Recursively flatten a nested sequence of elements.

Parameters:nested_list (iterable) – Possibly nested sequence of elements to flatten.
Returns:Flattened version of nested_list where any elements that are an iterable (collections.Iterable) have been unpacked into the top-level list, in a recursive fashion.
Return type:list
gensim.utils.getNS(host=None, port=None, broadcast=True, hmac_key=None)

Get a Pyro4 name server proxy.

Parameters:
  • host (str, optional) – Name server hostname.
  • port (int, optional) – Name server port.
  • broadcast (bool, optional) – Use broadcast mechanism? (i.e. reach out to all Pyro nodes in the network)
  • hmac_key (str, optional) – Private key.
Raises:

RuntimeError – When Pyro name server is not found.

Returns:

Proxy from Pyro4.

Return type:

Pyro4.core.Proxy

gensim.utils.get_max_id(corpus)

Get the highest feature id that appears in the corpus.

Parameters:corpus (iterable of iterable of (int, numeric)) – Collection of texts in BoW format.
Returns:Highest feature id.
Return type:int

Notes

For empty corpus return -1.

gensim.utils.get_my_ip()

Try to obtain our external ip (from the Pyro4 nameserver’s point of view)

Returns:IP address.
Return type:str

Warning

This tries to sidestep the issue of bogus /etc/hosts entries and other local misconfiguration, which often mess up hostname resolution. If all else fails, fall back to simple socket.gethostbyname() lookup.

gensim.utils.get_random_state(seed)

Generate numpy.random.RandomState based on input seed.

Parameters:seed ({None, int, array_like}) – Seed for random state.
Returns:Random state.
Return type:numpy.random.RandomState
Raises:AttributeError – If seed is not {None, int, array_like}.

Notes

Method originally from maciejkula/glove-python and written by @joshloyal.

gensim.utils.grouper(iterable, chunksize, as_numpy=False, dtype=<type 'numpy.float32'>)

Yield elements from iterable in “chunksize”-ed groups.

The last returned element may be smaller if the length of collection is not divisible by chunksize.

Parameters:
  • iterable (iterable of object) – An iterable.
  • chunksize (int) – Split iterable into chunks of this size.
  • as_numpy (bool, optional) – Yield chunks as np.ndarray instead of lists.
Yields:

list OR np.ndarray – “chunksize”-ed chunks of elements from iterable.

Examples

>>> print(list(grouper(range(10), 3)))
[[0, 1, 2], [3, 4, 5], [6, 7, 8], [9]]
gensim.utils.has_pattern()

Check whether the pattern package is installed.

Returns:Is pattern installed?
Return type:bool
gensim.utils.identity(p)

Identity fnc, for flows that don’t accept lambda (pickling etc).

Parameters:p (object) – Input parameter.
Returns:Same as p.
Return type:object
gensim.utils.is_corpus(obj)

Check whether obj is a corpus, by peeking at its first element. Works even on streamed generators. The peeked element is put back into a object returned by this function, so always use that returned object instead of the original obj.

Parameters:obj (object) – An iterable of iterable that contains (int, numeric).
Returns:Pair of (is obj a corpus, obj with peeked element restored)
Return type:(bool, object)

Examples

>>> from gensim.utils import is_corpus
>>> corpus = [[(1, 1.0)], [(2, -0.3), (3, 0.12)]]
>>> corpus_or_not, corpus = is_corpus(corpus)

Warning

An “empty” corpus (empty input sequence) is ambiguous, so in this case the result is forcefully defined as (False, obj).

gensim.utils.iter_windows(texts, window_size, copy=False, ignore_below_size=True, include_doc_num=False)

Produce a generator over the given texts using a sliding window of window_size.

The windows produced are views of some subsequence of a text. To use deep copies instead, pass copy=True.

Parameters:
  • texts (list of str) – List of string sentences.
  • window_size (int) – Size of sliding window.
  • copy (bool, optional) – Produce deep copies.
  • ignore_below_size (bool, optional) – Ignore documents that are not at least window_size in length?
  • include_doc_num (bool, optional) – Yield the text position with texts along with each window?
gensim.utils.keep_vocab_item(word, count, min_count, trim_rule=None)

Should we keep word in the vocab or remove it?

Parameters:
  • word (str) – Input word.
  • count (int) – Number of times that word appeared in a corpus.
  • min_count (int) – Discard words with frequency smaller than this.
  • trim_rule (function, optional) – Custom function to decide whether to keep or discard this word. If a custom trim_rule is not specified, the default behaviour is simply count >= min_count.
Returns:

True if word should stay, False otherwise.

Return type:

bool

gensim.utils.lazy_flatten(nested_list)

Lazy version of flatten().

Parameters:nested_list (list) – Possibly nested list.
Yields:object – Element of list
gensim.utils.lemmatize(content, allowed_tags=<_sre.SRE_Pattern object>, light=False, stopwords=frozenset([]), min_length=2, max_length=15)

Use the English lemmatizer from pattern to extract UTF8-encoded tokens in their base form aka lemma, e.g. “are, is, being” becomes “be” etc.

This is a smarter version of stemming, taking word context into account.

Parameters:
  • content (str) – Input string
  • allowed_tags (_sre.SRE_Pattern, optional) – Compiled regexp to select POS that will be used. Only considers nouns, verbs, adjectives and adverbs by default (=all other lemmas are discarded).
  • light (bool, optional) – DEPRECATED FLAG, DOESN’T SUPPORT BY pattern.
  • stopwords (frozenset, optional) – Set of words that will be removed from output.
  • min_length (int, optional) – Minimal token length in output (inclusive).
  • max_length (int, optional) – Maximal token length in output (inclusive).
Returns:

List with tokens with POS tags.

Return type:

list of str

Warning

This function is only available when the optional pattern is installed.

Raises:ImportError – If pattern not installed.

Examples

>>> from gensim.utils import lemmatize
>>> lemmatize('Hello World! How is it going?! Nonexistentword, 21')
['world/NN', 'be/VB', 'go/VB', 'nonexistentword/NN']

Note the context-dependent part-of-speech tags between these two examples:

>>> lemmatize('The study ranks high.')
['study/NN', 'rank/VB', 'high/JJ']
>>> lemmatize('The ranks study hard.')
['rank/NN', 'study/VB', 'hard/RB']
gensim.utils.merge_counts(dict1, dict2)

Merge dict1 of (word, freq1) and dict2 of (word, freq2) into dict1 of (word, freq1+freq2). :param dict1: First dictionary. :type dict1: dict of (str, int) :param dict2: Second dictionary. :type dict2: dict of (str, int)

Returns:result – Merged dictionary with sum of frequencies as values.
Return type:dict
gensim.utils.mock_data(n_items=1000, dim=1000, prob_nnz=0.5, lam=1.0)

Create a random Gensim-style corpus (BoW), using mock_data_row().

Parameters:
  • n_items (int) – Size of corpus
  • dim (int) – Dimension of vector, used for mock_data_row().
  • prob_nnz (float, optional) – Probability of each coordinate will be nonzero, will be drawn from Poisson distribution, used for mock_data_row().
  • lam (float, optional) – Parameter for Poisson distribution, used for mock_data_row().
Returns:

Gensim-style corpus.

Return type:

list of list of (int, float)

gensim.utils.mock_data_row(dim=1000, prob_nnz=0.5, lam=1.0)

Create a random gensim BoW vector, with the feature counts following the Poisson distribution.

Parameters:
  • dim (int, optional) – Dimension of vector.
  • prob_nnz (float, optional) – Probability of each coordinate will be nonzero, will be drawn from the Poisson distribution.
  • lam (float, optional) – Lambda parameter for the Poisson distribution.
Returns:

Vector in BoW format.

Return type:

list of (int, float)

gensim.utils.open_file(*args, **kwds)

Provide “with-like” behaviour without closing the file object.

Parameters:input (str or file-like) – Filename or file-like object.
Yields:file – File-like object based on input (or input if this already file-like).
gensim.utils.pickle(obj, fname, protocol=2)

Pickle object obj to file fname, using smart_open so that fname can be on S3, HDFS, compressed etc.

Parameters:
  • obj (object) – Any python object.
  • fname (str) – Path to pickle file.
  • protocol (int, optional) – Pickle protocol number. Default is 2 in order to support compatibility across python 2.x and 3.x.
gensim.utils.prune_vocab(vocab, min_reduce, trim_rule=None)

Remove all entries from the vocab dictionary with count smaller than min_reduce.

Modifies vocab in place, returns the sum of all counts that were pruned.

Parameters:
  • vocab (dict) – Input dictionary.
  • min_reduce (int) – Frequency threshold for tokens in vocab.
  • trim_rule (function, optional) – Function for trimming entities from vocab, default behaviour is vocab[w] <= min_reduce.
Returns:

result – Sum of all counts that were pruned.

Return type:

int

gensim.utils.pyro_daemon(name, obj, random_suffix=False, ip=None, port=None, ns_conf=None)

Register an object with the Pyro name server.

Start the name server if not running yet and block until the daemon is terminated. The object is registered under name, or name`+ some random suffix if `random_suffix is set.

gensim.utils.qsize(queue)

Get the (approximate) queue size where available.

Parameters:queue (queue.Queue) – Input queue.
Returns:Queue size, -1 if qsize method isn’t implemented (OS X).
Return type:int
gensim.utils.randfname(prefix='gensim')

Generate a random filename in temp.

Parameters:prefix (str) – Prefix of filename.
Returns:Full path in the in system’s temporary folder, ending in a random filename.
Return type:str
gensim.utils.revdict(d)

Reverse a dictionary mapping, i.e. {1: 2, 3: 4} -> {2: 1, 4: 3}.

Parameters:d (dict) – Input dictionary.
Returns:Reversed dictionary mapping.
Return type:dict

Notes

When two keys map to the same value, only one of them will be kept in the result (which one is kept is arbitrary).

Examples

>>> from gensim.utils import revdict
>>> d = {1: 2, 3: 4}
>>> revdict(d)
{2: 1, 4: 3}
gensim.utils.safe_unichr(intval)

Create a unicode character from its integer value. In case unichr fails, render the character as an escaped U<8-byte hex value of intval> string.

Parameters:intval (int) – Integer code of character
Returns:Unicode string of character
Return type:string
gensim.utils.sample_dict(d, n=10, use_random=True)

Selected n (possibly random) items from the dictionary d.

Parameters:
  • d (dict) – Input dictionary.
  • n (int, optional) – Number of items to select.
  • use_random (bool, optional) – Select items randomly (without replacement), instead of by the natural dict iteration order?
Returns:

Selected items from dictionary, as a list.

Return type:

list of (object, object)

gensim.utils.save_as_line_sentence(corpus, filename)

Save the corpus in LineSentence format, i.e. each sentence on a separate line, tokens are separated by space.

Parameters:corpus (iterable of iterables of strings) –
gensim.utils.simple_preprocess(doc, deacc=False, min_len=2, max_len=15)

Convert a document into a list of lowercase tokens, ignoring tokens that are too short or too long.

Uses tokenize() internally.

Parameters:
  • doc (str) – Input document.
  • deacc (bool, optional) – Remove accent marks from tokens using deaccent()?
  • min_len (int, optional) – Minimum length of token (inclusive). Shorter tokens are discarded.
  • max_len (int, optional) – Maximum length of token in result (inclusive). Longer tokens are discarded.
Returns:

Tokens extracted from doc.

Return type:

list of str

gensim.utils.simple_tokenize(text)

Tokenize input test using gensim.utils.PAT_ALPHABETIC.

Parameters:text (str) – Input text.
Yields:str – Tokens from text.
gensim.utils.smart_extension(fname, ext)

Append a file extension ext to fname, while keeping compressed extensions like .bz2 or .gz (if any) at the end.

Parameters:
  • fname (str) – Filename or full path.
  • ext (str) – Extension to append before any compression extensions.
Returns:

New path to file with ext appended.

Return type:

str

Examples

>>> from gensim.utils import smart_extension
>>> smart_extension("my_file.pkl.gz", ".vectors")
'my_file.pkl.vectors.gz'
gensim.utils.strided_windows(ndarray, window_size)

Produce a numpy.ndarray of windows, as from a sliding window.

Parameters:
  • ndarray (numpy.ndarray) – Input array
  • window_size (int) – Sliding window size.
Returns:

Subsequences produced by sliding a window of the given size over the ndarray. Since this uses striding, the individual arrays are views rather than copies of ndarray. Changes to one view modifies the others and the original.

Return type:

numpy.ndarray

Examples

>>> from gensim.utils import strided_windows
>>> strided_windows(np.arange(5), 2)
array([[0, 1],
       [1, 2],
       [2, 3],
       [3, 4]])
>>> strided_windows(np.arange(10), 5)
array([[0, 1, 2, 3, 4],
       [1, 2, 3, 4, 5],
       [2, 3, 4, 5, 6],
       [3, 4, 5, 6, 7],
       [4, 5, 6, 7, 8],
       [5, 6, 7, 8, 9]])
gensim.utils.synchronous(tlockname)

A decorator to place an instance-based lock around a method.

Notes

Adapted from http://code.activestate.com/recipes/577105-synchronization-decorator-for-class-methods/.

gensim.utils.to_unicode(text, encoding='utf8', errors='strict')

Convert text (bytestring in given encoding or unicode) to unicode.

Parameters:
  • text (str) – Input text.
  • errors (str, optional) – Error handling behaviour if text is a bytestring.
  • encoding (str, optional) – Encoding of text if it is a bytestring.
Returns:

Unicode version of text.

Return type:

str

gensim.utils.to_utf8(text, errors='strict', encoding='utf8')

Convert a unicode or bytes string in the given encoding into a utf8 bytestring.

Parameters:
  • text (str) – Input text.
  • errors (str, optional) – Error handling behaviour if text is a bytestring.
  • encoding (str, optional) – Encoding of text if it is a bytestring.
Returns:

Bytestring in utf8.

Return type:

str

gensim.utils.tokenize(text, lowercase=False, deacc=False, encoding='utf8', errors='strict', to_lower=False, lower=False)

Iteratively yield tokens as unicode strings, optionally removing accent marks and lowercasing it.

Parameters:
  • text (str or bytes) – Input string.
  • deacc (bool, optional) – Remove accentuation using deaccent()?
  • encoding (str, optional) – Encoding of input string, used as parameter for to_unicode().
  • errors (str, optional) – Error handling behaviour, used as parameter for to_unicode().
  • lowercase (bool, optional) – Lowercase the input string?
  • to_lower (bool, optional) – Same as lowercase. Convenience alias.
  • lower (bool, optional) – Same as lowercase. Convenience alias.
Yields:

str – Contiguous sequences of alphabetic characters (no digits!), using simple_tokenize()

Examples

>>> from gensim.utils import tokenize
>>> list(tokenize('Nic nemůže letět rychlostí vyšší, než 300 tisíc kilometrů za sekundu!', deacc=True))
[u'Nic', u'nemuze', u'letet', u'rychlosti', u'vyssi', u'nez', u'tisic', u'kilometru', u'za', u'sekundu']
gensim.utils.toptexts(*args, **kwargs)

Debug fnc to help inspect the top n most similar documents (according to a similarity index index), to see if they are actually related to the query.

Parameters:
  • query ({list of (int, number), numpy.ndarray}) – vector OR BoW (list of tuples)
  • texts (str) – object that can return something insightful for each document via texts[docid], such as its fulltext or snippet.
  • index (any) – A instance from from gensim.similarity.docsim.
Returns:

a list of 3-tuples (docid, doc’s similarity to the query, texts[docid])

Return type:

list

gensim.utils.trim_vocab_by_freq(vocab, topk, trim_rule=None)

Retain topk most frequent words in vocab. If there are more words with the same frequency as topk-th one, they will be kept. Modifies vocab in place, returns nothing.

Parameters:
  • vocab (dict) – Input dictionary.
  • topk (int) – Number of words with highest frequencies to keep.
  • trim_rule (function, optional) – Function for trimming entities from vocab, default behaviour is vocab[w] <= min_count.
gensim.utils.unpickle(fname)

Load object from fname, using smart_open so that fname can be on S3, HDFS, compressed etc.

Parameters:fname (str) – Path to pickle file.
Returns:Python object loaded from fname.
Return type:object
gensim.utils.upload_chunked(*args, **kwargs)

Memory-friendly upload of documents to a SimServer (or Pyro SimServer proxy).

Notes

Use this function to train or index large collections – avoid sending the entire corpus over the wire as a single Pyro in-memory object. The documents will be sent in smaller chunks, of chunksize documents each.