gistfile1.py

!sudo yum -y install java
!git clone https://www.github.com/interrogator/risk

import corpkit
from corpkit import interrogator, plotter, quickview
import pandas as pd
corpus = 'data/nyt/years'
#immediate sister to left of risk word
query = r'__ $. /(?i).?\brisk.?/'
# interrogate, output words only
precede = interrogator(corpus, 'words', query)
# now, repeat again
preceding_preceders = []
n = 5
for word in list(precede.results.columns)[:n]:
    query = r'__ $. /(?i).?\b%s.?/' % word
    interrogation = interrogator(corpus, 'words', query)
    preceding_preceders.append(interrogation.result)
results = pd.concat(preceding_preceders, axis = 1)
# then visualise your output somehow.