def read_embeddings(embeddings_filename, max_rank_emb):
"""Reads .vector or .bin file, modifies it to include <OOV> and <PADDING> and <SENTROOT>"""
if embeddings_filename.endswith(".bin"):
binary = True
else:
binary = False
gensim_vectors = KeyedVectors.load_word2vec_format(
embeddings_filename,
binary=binary,
limit=max_rank_emb,
unicode_errors='replace')
gensim_vectors.vocab["<OOV>"] = Vocab(index=1)
gensim_vectors.vocab["<PADDING>"] = Vocab(index=0)
gensim_vectors.vocab["<SENTROOT>"] = Vocab(index=2)
for word_record in gensim_vectors.vocab.values():
word_record.index += 3
two_random_rows = numpy.random.uniform(
low=-0.01, high=0.01, size=(3, gensim_vectors.vectors.shape[1]))
# stack the two rows, and the embedding matrix on top of each other
gensim_vectors.vectors = numpy.vstack(
[two_random_rows, gensim_vectors.vectors])
gensim_vectors.vectors = keras.utils.normalize(gensim_vectors.vectors,
axis=0)
gensim_vectors.vectors = keras.utils.normalize(gensim_vectors.vectors)
return gensim_vectors
def add_word(word, weights):
word_id = len(result.vocab)
if word in result.vocab:
return
result.vocab[word] = Vocab(index=word_id, count=vocab_size - word_id)
result.syn0[word_id] = weights
result.index2word.append(word)
def create_keyedvector_from_matrix(self, embedding_matrix, word2id):
"""
Imports the necessary attributes for the Embedding object from an embedding matrix and a word2id vocabulary. Can be used for custom pre-trained embeddings.
Parameters
----------
embedding_matrix: numpy.ndarray
Embedding matrix as a numpy object
word2id: dict
Word vocabulary (key: word, value: word_index)
"""
vocab = {
word: word2id[word]
for word in sorted(word2id, key=word2id.__getitem__, reverse=False)
}
embedding_matrix = embedding_matrix
vector_size = embedding_matrix.shape[1]
kv = KeyedVectors(vector_size)
kv.vector_size = vector_size
kv.vectors = embedding_matrix
kv.index2word = list(vocab.keys())
kv.vocab = {
word: Vocab(index=word_id, count=0)
for word, word_id in vocab.items()
}
self.embedding = kv
def __init__(self,
model_path,
model_params_path,
dim,
dtype="float32",
dtype_size=4,
header_size=32):
self.header_size_bites = header_size * dtype_size
self.dtype_size = dtype_size
self.dtype = dtype
self.dim = dim
with open(model_params_path) as fd:
params = json.load(fd)
self.model = gensim.models.keyedvectors.FastTextKeyedVectors(
vector_size=dim,
min_n=params['hash_params']['minn'],
max_n=params['hash_params']['maxn'],
bucket=params['hash_params']['num_buckets'],
compatible_hash=params['hash_params']['fb_compatible'])
self.model.vectors_vocab = self.load_matrix(
f'{model_path}.vectors_vocab.npy')
self.model.vectors = self.load_matrix(f'{model_path}.vectors.npy')
self.model.vectors_ngrams = self.load_matrix(
f'{model_path}.vectors_ngrams.npy')
self.model.vocab = dict((word, Vocab(index=idx, count=1))
for word, idx in params['vocab'].items())
def _load_tree(self, tree: dict) -> None:
self.__dict__.update(tree)
self.tokens = split_strings(self.tokens)
self.frequencies = {
w: self.frequencies["vals"][i]
for i, w in enumerate(split_strings(self.frequencies["keys"]))}
self.checker = SymSpell(max_dictionary_edit_distance=self.max_distance)
self.checker.__dict__.update(tree["checker"])
deletes = {}
words = split_strings(self.checker._deletes["strings"])
lengths = self.checker._deletes["lengths"]
data = self.checker._deletes["data"]
offset = 0
for i, delindex in enumerate(self.checker._deletes["indexes"]):
length = lengths[i]
deletes[delindex] = [words[j] for j in data[offset:offset + length]]
offset += length
self.checker._deletes = deletes
self.checker._words = {w: self.checker._words[i] for i, w in enumerate(words)}
vectors = self.wv["vectors"]
wv = FastTextKeyedVectors(vectors.shape[1], self.wv["min_n"], self.wv["max_n"],
self.wv["bucket"], True)
wv.vectors = numpy.array(vectors)
vocab = split_strings(self.wv["vocab"]["strings"])
wv.vocab = {
s: Vocab(index=i, count=self.wv["vocab"]["counts"][i])
for i, s in enumerate(vocab)}
wv.bucket = self.wv["bucket"]
wv.index2word = wv.index2entity = vocab
wv.num_ngram_vectors = self.wv["num_ngram_vectors"]
wv.vectors_ngrams = numpy.array(self.wv["vectors_ngrams"])
wv.hash2index = {k: v for v, k in enumerate(self.wv["hash2index"])}
self.wv = wv
def load_vocab(wv, vocab_file, use_glove_format, restrict_vocab):
# Read vocab.
vocab_size = 0
with open(vocab_file, "r") as f:
wv.index2freq = []
all_lines = f.readlines()[:restrict_vocab] if restrict_vocab > 0 else f.readlines()
for index, line in enumerate(all_lines):
if use_glove_format:
word, count = line.strip().split(" ") # vocab is indexed from 0; for co-occ we use 1-based indexing
index = index
else:
index, word, count = line.strip().split("\t")
index = int(index) - 1 # indexing starts at 1 in the file; for co-occ we use 0-based indexing
wv.index2word.append(word)
wv.vocab[word] = Vocab(index=index, count=int(count))
wv.index2freq.append(count)
vocab_size += 1
wv.index2freq = array(wv.index2freq).astype(uint32)
# Unused members from VanillaWordEmbeddingsKeyedVectors.
wv.vectors_norm = None
print("Loaded vocabulary with {} words".format(vocab_size))
return vocab_size
def convert_graph_embedding_to_gensim(
model_path: Union[str, Path],
dataset_path: Union[str, Path],
embedding_path: Union[str, Path] = None) -> Path:
"""
Example usage
convert_graph_embedding_to_gensim(
model_path=(
DEFAULT_OUTPUT_PATH / 'reviews_Cell_Phones_and_Accessories-50000-docs' / 'our' / 'aspect_2_aspect_graph-en_core_web_lg.en_core_web_lg.model').as_posix(),
dataset_path=(
DEFAULT_OUTPUT_PATH / 'reviews_Cell_Phones_and_Accessories-50000-docs' / 'our' / 'aspect_2_aspect_graph-en_core_web_lg.en_core_web_lg.dataset').as_posix()
)
"""
if embedding_path is None:
embedding_path = Path(model_path).with_suffix('.word2vec_format.bin')
model = torch.load(model_path)
dataset = torch.load(dataset_path)
_save_word2vec_format(
embedding_path,
vocab={
word.replace(' ', '_'): Vocab(index=index, count=1)
for word, index in dataset[0].nodes_mapping[0].items()
},
vectors=model.embedding.weight.detach().numpy(),
binary=True)
return embedding_path
def add_word(word, weights):
word_id = len(result.vocab)
if word in result.vocab:
logger.warning("duplicate word '%s' in %s, ignoring all but first", word, fname)
return
if counts is None:
# most common scenario: no vocab file given. just make up some bogus counts, in descending order
result.vocab[word] = Vocab(index=word_id, count=vocab_size - word_id)
elif word in counts:
# use count from the vocab file
result.vocab[word] = Vocab(index=word_id, count=counts[word])
else:
# vocab file given, but word is missing -- set count to None (TODO: or raise?)
logger.warning("vocabulary file is incomplete: '%s' is missing", word)
result.vocab[word] = Vocab(index=word_id, count=None)
result.vectors[word_id] = weights
result.index2word.append(word)
def generate_vocab(walks):
index2word = []
raw_vocab = defaultdict(int)
for walk in walks:
for word in walk:
raw_vocab[word] += 1
vocab = {}
for word, v in iteritems(raw_vocab):
vocab[word] = Vocab(count=v, index=len(index2word))
index2word.append(word)
index2word.sort(key=lambda word: vocab[word].count, reverse=True)
for i, word in enumerate(index2word):
vocab[word].index = i
return vocab, index2word
def add_embeddings(keyed_vectors, *words, init=None):
from gensim.models.keyedvectors import Vocab
if init is None:
init = np.zeros
syn0 = keyed_vectors.syn0
for word in words:
keyed_vectors.key_to_index[word] = Vocab(count=0, index=syn0.shape[0])
keyed_vectors.syn0 = np.concatenate([syn0, init((1, syn0.shape[1]))])
keyed_vectors.index2word.append(word)
return syn0.shape[0]
def save_gensim_model(words, word_reprs, output_file, binary=True):
"""Save word representations in w2v format. Word order is not preserved"""
vocab = dict()
for word in words:
vocab[word] = Vocab(index=len(vocab))
model = Word2VecKeyedVectors(word_reprs.shape[1])
model.vocab = vocab
model.vectors = word_reprs
model.save_word2vec_format(fname=output_file, binary=binary)
def generate_vocab(all_walks):
index2word = []
raw_vocab = defaultdict(int)
for walks in all_walks: # 遍历每一层walks
for walk in walks:
for word in walk:
raw_vocab[word] += 1 # 统计一下word（node）出现的次数
vocab = {}
for word, v in iteritems(raw_vocab): # 构建词袋，给word一个编码，按照出现的次数进行排序
vocab[word] = Vocab(count=v, index=len(index2word))
index2word.append(word)
index2word.sort(key=lambda word: vocab[word].count,
reverse=True) # 按照词袋的数量从大到小进行排序
for i, word in enumerate(index2word):
vocab[word].index = i # 词袋的index重新排序
return vocab, index2word
def load(root_model):
with open(root_model + '.txt', encoding='utf-8') as f:
index_vocab = [t.strip() for t in f]
vocab = {w: Vocab(index=i) for (i, w) in enumerate(index_vocab)}
vectors = np.load(root_model + '.npy')
m = Word2Vec()
m.wv.index2word = index_vocab
m.wv.vectors = vectors
m.wv.vocab = vocab
return m
def add_embeddings(keyed_vectors, *words, init=None):
import numpy as np
from gensim.models.keyedvectors import Vocab
if init is None:
init = np.zeros
vectors = keyed_vectors.vectors
for word in words:
keyed_vectors.vocab[word] = Vocab(count=0, index=vectors.shape[0])
keyed_vectors.vectors = np.concatenate(
[vectors, init((1, vectors.shape[1]))])
keyed_vectors.index2word.append(word)
return vectors.shape[0]
def add(self, wwc: Collection[Tuple[str, np.ndarray, int]]) -> Iterator[Vocab]:
"""`wwc` is a collection of (word, weights, count) tuples"""
vocab_size, vector_size = self.vectors.shape
self.vectors.resize((vocab_size + len(wwc), vector_size))
for i, (word, weights, count) in enumerate(wwc):
word_id = vocab_size + i
voc = Vocab(index=word_id, count=count)
self.vocab[word] = voc
self.vectors[word_id] = weights
self.index2word.append(word)
yield voc
def generate_vocab(all_walks):
index2word = []
raw_vocab = defaultdict(int)
for walks in all_walks:
for walk in walks:
for word in walk:
raw_vocab[word] += 1
vocab = {}
for word, v in iteritems(raw_vocab):
vocab[word] = Vocab(count=v, index=len(index2word))
index2word.append(word)
# 按照每个单词出现的频次进行从大到小排序在vocab中
index2word.sort(key=lambda word: vocab[word].count, reverse=True)
for i, word in enumerate(index2word):
vocab[word].index = i
# vocab是一个按照walks中所有节点出现的频率从大到小排序后的单词表
# index2word是节点集合
return vocab, index2word
def _set_keyedvector(self, attrname, keys, dim, vec=None):
keyed_vec = KeyedVectors(dim)
dummy_max_count = len(keys) + 1
for i, key in enumerate(keys):
key = str(key)
keyed_vec.vocab[key] = Vocab(index=i, count=dummy_max_count - i) # dummy count
keyed_vec.index2word.append(key)
if vec is not None:
keyed_vec.vectors = vec
keyed_vec.init_sims()
setattr(self, attrname, keyed_vec)
def load_emb(fp):
n_vocab, dim = map(int, fp.readline().split())
emb = KeyedVectors(dim)
emb.vectors = np.empty((n_vocab, dim), dtype=np.float32)
for i, line in tqdm(enumerate(fp), total=n_vocab, unit='word'):
word, vec_str = line.split(' ', 1)
emb.vectors[i] = np.fromstring(vec_str, sep=' ')
emb.vocab[word] = Vocab(index=i)
emb.index2word.append(word)
return emb
def _create_pv(self):
try:
vectors = KeyedVectors()
except TypeError:
# Newer versions of gensim require a constructor argument.
vectors = KeyedVectors(self._wv.shape[1])
phrases = []
for name in self.terminology.iter_names():
# This iterates over unique names.
vectors.vocab[name] = Vocab(index=len(vectors.vocab), count=None)
vectors.index2word.append(name)
phrases.append(self._phrase_vector(name))
vectors.syn0 = vectors.syn0norm = np.array(phrases)
return vectors
def generate_vocab(all_walks):
index2word = []
raw_vocab = defaultdict(
int) # stores the count of a word appearing in the walk
for walks in all_walks:
for walk in walks:
for word in walk:
raw_vocab[word] += 1
# compute count, then to sort based on count, at last set index after sort
vocab = {}
for word, v in iteritems(raw_vocab):
vocab[word] = Vocab(
count=v, index=len(index2word)) # vocab[word] = (count, index)
index2word.append(word) # index2word[index] == word
index2word.sort(key=lambda word: vocab[word].count,
reverse=True) # decending order
for i, word in enumerate(index2word):
vocab[word].index = i # word2inedx
return vocab, index2word # vocab:
def update(self):
wv = self.word_vectors_file.get_word_vectors()
voc = self.vocabs_file.get_vocabs()['word']
words_in_vocab = [
k for k, _ in sorted(voc.items(), key=lambda i: i[1][0])
]
word_embs = wv[words_in_vocab[1:]]
unk_emb = np.mean(word_embs, 0, keepdims=True)
embs = np.concatenate((unk_emb, word_embs), 0)
kv = KeyedVectors(embs.shape[1])
kv.syn0 = embs
kv.vocab = dict(
(k, Vocab(index=v[0], count=v[1])) for k, v in voc.items())
kv.index2word = words_in_vocab
kv.save(self.path)
def __create_keyed_vector(matrix, orig_vocab):
vocab = dict()
index_to_word = []
for word, word_id in sorted(orig_vocab.token2id.items(),
key=itemgetter(1)):
index_to_word.append(word)
vocab[word] = Vocab(index=word_id, count=orig_vocab.word_freq[word_id])
vector_size = matrix.shape[1]
keyed_vector = KeyedVectors(vector_size)
keyed_vector.vector_size = vector_size
keyed_vector.vocab = vocab
keyed_vector.index2word = index_to_word
keyed_vector.vectors = matrix
assert (len(vocab), vector_size) == keyed_vector.vectors.shape
return keyed_vector
def _load_dict(self, file_handle, encoding='utf8'):
"""Load a previously saved dictionary from disk, stored in Facebook's native fasttext format.
Parameters
----------
file_handle : file-like object
The opened file handle to the persisted dictionary.
encoding : str
Specifies the encoding.
"""
vocab_size, nwords, nlabels = self.struct_unpack(file_handle, '@3i')
# Vocab stored by [Dictionary::save](https://github.com/facebookresearch/fastText/blob/master/src/dictionary.cc)
if nlabels > 0:
raise NotImplementedError("Supervised fastText models are not supported")
logger.info("loading %s words for fastText model from %s", vocab_size, self.file_name)
self.struct_unpack(file_handle, '@1q') # number of tokens
if self.new_format:
pruneidx_size, = self.struct_unpack(file_handle, '@q')
for i in range(vocab_size):
word_bytes = b''
char_byte = file_handle.read(1)
# Read vocab word
while char_byte != b'\x00':
word_bytes += char_byte
char_byte = file_handle.read(1)
word = word_bytes.decode(encoding)
count, _ = self.struct_unpack(file_handle, '@qb')
self.wv.vocab[word] = Vocab(index=i, count=count)
self.wv.index2word.append(word)
assert len(self.wv.vocab) == nwords, (
'mismatch between final vocab size ({} words), '
'and expected number of words ({} words)'.format(len(self.wv.vocab), nwords))
if len(self.wv.vocab) != vocab_size:
# expecting to log this warning only for pretrained french vector, wiki.fr
logger.warning(
"mismatch between final vocab size (%s words), and expected vocab size (%s words)",
len(self.wv.vocab), vocab_size
)
if self.new_format:
for j in range(pruneidx_size):
self.struct_unpack(file_handle, '@2i')
def load_dict(self, file_handle, encoding='utf8'):
vocab_size, nwords, _ = self.struct_unpack(file_handle, '@3i')
# Vocab stored by [Dictionary::save](https://github.com/facebookresearch/fastText/blob/master/src/dictionary.cc)
logger.info("loading %s words for fastText model from %s", vocab_size,
self.file_name)
self.struct_unpack(file_handle, '@1q') # number of tokens
if self.new_format:
pruneidx_size, = self.struct_unpack(file_handle, '@q')
for i in range(vocab_size):
word_bytes = b''
char_byte = file_handle.read(1)
# Read vocab word
while char_byte != b'\x00':
word_bytes += char_byte
char_byte = file_handle.read(1)
word = word_bytes.decode(encoding)
count, _ = self.struct_unpack(file_handle, '@qb')
if i == nwords and i < vocab_size:
# To handle the error in pretrained vector wiki.fr (French).
# For more info : https://github.com/facebookresearch/fastText/issues/218
assert word == "__label__", (
'mismatched vocab_size ({}) and nwords ({}), extra word "{}"'
.format(vocab_size, nwords, word))
continue # don't add word to vocab
self.wv.vocab[word] = Vocab(index=i, count=count)
self.wv.index2word.append(word)
assert len(self.wv.vocab) == nwords, (
'mismatch between final vocab size ({} words), '
'and expected number of words ({} words)'.format(
len(self.wv.vocab), nwords))
if len(self.wv.vocab) != vocab_size:
# expecting to log this warning only for pretrained french vector, wiki.fr
logger.warning(
"mismatch between final vocab size (%s words), and expected vocab size (%s words)",
len(self.wv.vocab), vocab_size)
if self.new_format:
for j in range(pruneidx_size):
self.struct_unpack(file_handle, '@2i')
def load_a_format(cls,
fin,
vocab_size,
vector_size,
datatype=np.float32,
discard=None):
# type: (Iterable[str], int, int, Any, Optional[Callable[[str], bool]]) -> KeyedVectorsOriginal
discard = discard or false
result = cls(vector_size)
result.vector_size = vector_size
result.vectors = np.zeros((vocab_size, vector_size), dtype=datatype)
for line in fin:
word, vect = line.rstrip().split(" ", 1)
if discard(word):
continue
weights = np.fromstring(vect, sep=" ", dtype=np.float32)
# raise ValueError("invalid vector on line %s; is vector_size incorrect or file otherwise damaged?" % (i+1,))
if word in result.vocab:
raise DuplicateEntry(word)
word_id = len(result.index2word)
result.vocab[word] = Vocab(index=word_id,
count=vocab_size - word_id)
result.vectors[word_id] = weights
result.index2word.append(word)
real_size = len(result.index2word)
if real_size != vocab_size:
if discard is None:
raise EOFError(
"unexpected end of input; is vocab_size incorrect or file otherwise damaged?"
)
else:
result.vectors.resize(
(real_size, vector_size)) # this should be no-copy
return result
def prep_embeddings_fast(ft, emb_func, limit_vocab_size=30000):
w2v = Word2Vec()
emb = None
if type(emb_func) is np.ndarray:
size = min(limit_vocab_size, emb_func.shape[0])
emb = emb_func[:size]
else:
emb = np.zeros((limit_vocab_size, ft.D))
for i in range(limit_vocab_size):
emb[i], _ = emb_func(ft.id2word[i])
size = emb.shape[0]
w2v.index2word = ft.id2word[:size]
w2v.vector_size = ft.D
w2v.syn0 = emb
dvocab = {}
for word_id, word in enumerate(w2v.index2word):
dvocab[word] = Vocab(index=word_id, count=ft.nwords - word_id)
w2v.vocab = dvocab
return w2v
def main():
"""Entry point."""
parser = argparse.ArgumentParser("AWD-LSTM Embeddings to Word Vectors")
parser.add_argument("--model", required=True)
parser.add_argument("--dictionary", required=True)
parser.add_argument("--output", required=True)
args = parser.parse_args()
dictionary = torch.load(args.dictionary)
model = torch.load(args.model, map_location='cpu')
embeddings = model[0].encoder.weight.data.cpu().numpy()
kv = KeyedVectors(embeddings.shape[1])
kv.syn0 = embeddings
kv.vocab = {
w: Vocab(index=i)
for i, w in enumerate(dictionary.dictionary.idx2word)
}
kv.index2word = dictionary.dictionary.idx2word
kv.save(args.output)
def prepare_vocab(self, hs, negative, wv, update=False,
keep_raw_vocab=False, trim_rule=None,
min_count=None, sample=None, dry_run=False):
min_count = min_count or self.min_count
sample = sample or self.sample
drop_total = drop_unique = 0
if not update:
raise Exception('Nonce2Vec can only update a pre-existing '
'vocabulary')
logger.info('Updating model with new vocabulary')
new_total = pre_exist_total = 0
# New words and pre-existing words are two separate lists
new_words = []
pre_exist_words = []
if self.nonce is not None:
# if self.nonce is not None and self.nonce in wv.vocab:
if self.nonce in wv.vocab:
gold_nonce = '{}_true'.format(self.nonce)
nonce_index = wv.vocab[self.nonce].index
wv.vocab[gold_nonce] = wv.vocab[self.nonce]
wv.index2word[nonce_index] = gold_nonce
# del wv.index2word[wv.vocab[self.nonce].index]
del wv.vocab[self.nonce]
for word, v in iteritems(self.raw_vocab):
# Update count of all words already in vocab
if word in wv.vocab:
pre_exist_words.append(word)
pre_exist_total += v
if not dry_run:
wv.vocab[word].count += v
else:
# For new words, keep the ones above the min count
# AND the nonce (regardless of count)
if keep_vocab_item(word, v, min_count,
trim_rule=trim_rule) or word == self.nonce:
new_words.append(word)
new_total += v
if not dry_run:
wv.vocab[word] = Vocab(count=v,
index=len(wv.index2word))
wv.index2word.append(word)
else:
drop_unique += 1
drop_total += v
original_unique_total = len(pre_exist_words) \
+ len(new_words) + drop_unique
pre_exist_unique_pct = len(pre_exist_words) \
* 100 / max(original_unique_total, 1)
new_unique_pct = len(new_words) * 100 / max(original_unique_total, 1)
logger.info('New added %i unique words (%i%% of original %i) '
'and increased the count of %i pre-existing words '
'(%i%% of original %i)', len(new_words),
new_unique_pct, original_unique_total,
len(pre_exist_words), pre_exist_unique_pct,
original_unique_total)
retain_words = new_words + pre_exist_words
retain_total = new_total + pre_exist_total
# Precalculate each vocabulary item's threshold for sampling
if not sample:
# no words downsampled
threshold_count = retain_total
# Only retaining one subsampling notion from original gensim implementation
else:
threshold_count = sample * retain_total
downsample_total, downsample_unique = 0, 0
for w in retain_words:
v = wv.vocab[w].count
word_probability = (np.sqrt(v / threshold_count) + 1) \
* (threshold_count / v)
if word_probability < 1.0:
downsample_unique += 1
downsample_total += word_probability * v
else:
word_probability = 1.0
downsample_total += v
if not dry_run:
wv.vocab[w].sample_int = int(round(word_probability * 2**32))
if not dry_run and not keep_raw_vocab:
logger.info('deleting the raw counts dictionary of %i items',
len(self.raw_vocab))
self.raw_vocab = defaultdict(int)
logger.info('sample=%g downsamples %i most-common words', sample,
downsample_unique)
logger.info('downsampling leaves estimated %i word corpus '
'(%.1f%% of prior %i)', downsample_total,
downsample_total * 100.0 / max(retain_total, 1),
retain_total)
# return from each step: words-affected, resulting-corpus-size,
# extra memory estimates
report_values = {
'drop_unique': drop_unique, 'retain_total': retain_total,
'downsample_unique': downsample_unique,
'downsample_total': int(downsample_total),
'num_retained_words': len(retain_words)
}
if self.null_word:
# create null pseudo-word for padding when using concatenative
# L1 (run-of-words)
# this word is only ever input – never predicted – so count,
# huffman-point, etc doesn't matter
self.add_null_word(wv)
if self.sorted_vocab and not update:
self.sort_vocab(wv)
if hs:
# add info about each word's Huffman encoding
self.create_binary_tree(wv)
if negative:
# build the table for drawing random words (for negative sampling)
self.make_cum_table(wv)
return report_values, pre_exist_words
