Formatting scraped ICLR Reviews

format_scraped_iclr_reviews.py

url = 'https://openreview.net/notes?invitation=ICLR.cc%2F2018%2FConference%2F-%2FBlind_Submission&offset=0&limit=1000'
df = pd.DataFrame(requests.get(url).json()['notes']) # Each row in this data frame is a paper.
forum_content = []
for i, forum_id in list(enumerate(df.forum)): # Each forum_id is a review, comment, or acceptance decision about a paper.
  forum_content.append(requests.get('https://openreview.net/notes?forum={}&trash=true'.format(forum_id)).json())
  time.sleep(.3)
df['forumContent'] = pd.Series(forum_content)

df['title'] = df.content.apply(lambda x: x['title'])
df['authors'] = df.content.apply(lambda x: x['authors'])
df['decision_raw'] = df.forumContent.apply(lambda x:[n['content']['decision'] for n in x['notes'] 
                                                     if 'decision' in n['content']][0])
'''
>>> df['decision_raw'].value_counts()
Reject                      504
Accept (Poster)             313
Invite to Workshop Track     90
Accept (Oral)                23
Name: decision_raw, dtype: int64
'''

# Create a data frame of reviews from forum content.  Note that the colummn "forum" is also present in df
only_reviews_df = pd.concat(df.forumContent.apply(lambda c: pd.DataFrame([
    {'review': n['content']['review'], 
     'rating': n['content']['rating'],  
     'confidence': n['content']['confidence'],
     'forum': n['forum']} 
    for n in c['notes'] 
    if 'content' in n and 'review' in n['content']
])).tolist())

# merge it with the paper and decision information
reviews_df = pd.merge(df[['title', 'authors', 'decision_raw', 'forum']], only_reviews_df, on='forum')
'''
title           Improving Discriminator-Generator Balance in G...
authors                         [Simen Selseng and Björn Gambäck]
decision_raw                                               Reject
forum                                                   SyBPtQfAZ
confidence      4: The reviewer is confident but not absolutel...
rating                                         3: Clear rejection
review          The paper proposes a variety of modifications ...
Name: 0, dtype: object
Note: there are 2806 reviews
'''

# Categorize the reviews by Accept (>6), Neutral (4, 5, and 6), and Reject (<4),
# and categorize the acceptance decisions as Accept (paper or poster), Workshop, or Reject. 
# Concatentate them together into a single category.
reviews_df['decision'] = (reviews_df['decision_raw']
                          .apply(lambda x: 'Reject' if x == 'Reject' 
                                 else ('Accept' if x.startswith('Accept') 
                                       else 'Workshop')))
reviews_df['rating_bin'] = (reviews_df['rating']
                            .apply(lambda x: (lambda s: 'Negative' if s < 5 
                                              else ('Positive' if s > 6 else 'Neutral'))
                                   (int(x.split(':')[0].strip()))))
reviews_df['category'] = reviews_df['decision'] + ', ' + reviews_df['rating_bin']
reviews_df.to_csv('iclr2018_reviews.csv.bz2', index=False, compression='bz2')