Biterm Topic Model
Project description
Biterm Topic Model
This is a simple Cython implenentaion of the awesome Biterm Topic Model. This model is accurate in short text classification. It explicitly models the word co-occurrence patterns in the whole corpus to solve the problem of sparse word co-occurrence at document-level.
Simply install by:
pip install biterm
Load some short texts and vectorize them via sklearn.
from sklearn.feature_extraction.text import CountVectorizer
texts = open('./data/reuters.titles').read().splitlines()[:50]
vec = CountVectorizer(stop_words='english')
X = vec.fit_transform(texts).toarray()
Get the vocabulary and the biterms from the texts.
from biterm.utility import vec_to_biterms
vocab = np.array(vec.get_feature_names())
biterms = vec_to_biterms(X)
Create a BTM and pass the biterms to train it.
from biterm.cbtm import oBTM
btm = oBTM(num_topics=20, V=vocab)
topics = btm.fit_transform(biterms, iterations=100)
Save a topic plot using pyLDAvis and explore the results! (also see simple_btml.py)
from biterm.btm import oBTM
btm = oBTM(num_topics=20, V=vocab)
topics = btm.fit_transform(biterms, iterations=100)
Inference is done with Gibbs Sampling and it's not really fast. The implementation is not meant for production. But if you have to classify a lot of texts you can try using online learning. Also use the Cython version.
import numpy as np
import pyLDAvis
from biterm.cbtm import oBTM # use Cython model
from sklearn.feature_extraction.text import CountVectorizer
from biterm.utility import vec_to_biterms, topic_summuary # helper functions
if __name__ == "__main__":
texts = open('./data/reuters.titles').read().splitlines()
# vectorize texts
vec = CountVectorizer(stop_words='english')
X = vec.fit_transform(texts).toarray()
# get vocabulary
vocab = np.array(vec.get_feature_names())
# get biterms
biterms = vec_to_biterms(X)
# create btm
btm = oBTM(num_topics=20, V=vocab)
print("\n\n Train Online BTM ..")
for i in range(0, len(biterms), 100): # prozess chunk of 200 texts
biterms_chunk = biterms[i:i + 100]
btm.fit(biterms_chunk, iterations=50)
topics = btm.transform(biterms)
print("\n\n Visualize Topics ..")
vis = pyLDAvis.prepare(btm.phi_wz.T, topics, np.count_nonzero(X, axis=1), vocab, np.sum(X, axis=0))
pyLDAvis.save_html(vis, './vis/online_btm.html')
print("\n\n Topic coherence ..")
topic_summuary(btm.phi_wz.T, X, vocab, 10)
print("\n\n Texts & Topics ..")
for i in range(len(texts)):
print("{} (topic: {})".format(texts[i], topics[i].argmax()))
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for biterm-0.1.1-cp36-cp36m-macosx_10_7_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8ad62c810d0e240691faf2cb6b1f720d40a7436241e4f6b9bc8b05b04d72223b |
|
MD5 | 5ee7d5e94828ee65c8c1924ff92f1a08 |
|
BLAKE2b-256 | ced6383ac811cd3ed4e1faf331cfc6896bc153a36b84632ec794a54c633809b0 |