Quick n-gram discovery on the NIPS 2016 accepted papers, from https://nips.cc/Conferences/2016/AcceptedPapers.

nips-accepted-papers.py

from sklearn.feature_extraction.text import CountVectorizer
from collections import Counter
import numpy as np

with open('accepted-papers.txt', 'r') as myfile:
	data=myfile.read().replace('\n', '')

lectures = data
bigram_vectorizer = CountVectorizer(ngram_range=(2, 10), stop_words='english')
analyze = bigram_vectorizer.build_analyzer()
output = analyze(lectures)

print(Counter(output).most_common(50))

# [(u'stanford university', 54), (u'carnegie mellon', 45), (u'mellon university', 44), (u'carnegie mellon university', 44), (u'google deepmind', 43), (u'princeton university', 37), (u'uc berkeley', 35), (u'neural networks', 28), (u'microsoft research', 25), (u'university washington', 24), (u'cornell university', 18), (u'eth zurich', 17), (u'university michigan', 16), (u'university texas', 16), (u'university pennsylvania', 16), (u'university toronto', 15), (u'columbia university', 15), (u'duke university', 14), (u'tsinghua university', 14), (u'university oxford', 14), (u'university texas austin', 13), (u'texas austin', 13), (u'state university', 13), (u'university montreal', 12), (u'ut austin', 12), (u'university cambridge', 11), (u'peking university', 11), (u'university california', 10), (u'university minnesota', 10), (u'stochastic gradient', 10), (u'university wisconsin', 9), (u'institute science', 9), (u'deep learning', 8), (u'national university', 8), (u'university chicago', 8), (u'bayesian optimization', 8), (u'convolutional neural', 8), (u'convex optimization', 8), (u'harvard university', 8), (u'virginia tech', 7), (u'national laboratory', 7), (u'san diego', 7), (u'chen university', 7), (u'supervised learning', 7), (u'alamos national', 7), (u'alamos national laboratory', 7), (u'new york', 7), (u'recurrent neural networks', 7), (u'los alamos national laboratory', 7), (u'university david', 7)]