Update distillability modeling scripts to predict long articles

heuristics/distillable/write_features_csv.py

 #!/usr/bin/env python
 # Copyright 2016 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 import argparse
 import csv
 import json
 import os
 import shutil
 import sys
+import unittest
+
+def filter_fields(header, data, fields):
+  """Filter (header, data) with selected header fields.
+
+  Args:
+    header ([str]): The header.
+    data ([[float]]): The data, with the same number of columns as header.
+    fields ([str]): The fields that need to be written.
+
+  Returns:
+    (header ([str]), data ([[float]]))
+
+  Examples:
+    >>> filter_fields(['a','b','c'], [[0,1,2], [3,4,5]], ['d','b','a'])
+    (['b', 'a'], [[1, 0], [4, 3]])
+  """
+
+  picked = []
+  for f in fields:
+    try:
+      picked.append(header.index(f))
+    except ValueError:
+      # OK to have missing values
+      pass
+
+  h = [header[i] for i in picked]
+  d = []
+  for e in data:
+    d.append([e[i] for i in picked])
+  return (h, d)
+
+def write_features(filename, header, data, fields):
+  """Write (header, data) to filename in CSV format, with selected fields.
+
+  Args:
+    filename (str): The output filename.
+    header ([str]): The header.
+    data ([[float]]): The data, with the same number of columns as header.
+    fields ([str]): The fields that need to be written.
+
+  Examples:
+    >>> write_features(None, ['a','b','c'], [[0,1,2], [3,4,5]], ['d','b','a'])
+    b,a
+    1,0
+    4,3
+  """
+
+  (header, data) = filter_fields(header, data, fields)
+
+  if filename:
+    writer = csv.writer(open(filename, 'w'))
+  else:
+    writer = csv.writer(sys.stdout, lineterminator="\n")
+
+  writer.writerow(header)
+  writer.writerows(data)
+
+def getGroups(header):
+  """Return groups of header fields
+
+  Returns:
+    dict of name (str): fields ([str])
+  """
+  groups = {}
+  groupPath = [
+    'forum',
+    'index',
+    'search',
+    'view',
+    'archive',
+    'asp',
+    'phpbb',
+    'php',
+    'pathLength',
+    'domain',
+    'pathComponents',
+    'slugDetector',
+    'pathNumbers',
+    'lastSegmentLength',
+  ]
+  groups['path'] = groupPath
+
+  groupNumElement = [
+    'numElements',
+    'numAnchors',
+    'anchorRatio',
+    'numForms',
+    'numTextInput',
+    'numPasswordInput',
+    'numPPRE',
+  ]
+  groups['numElement'] = groupNumElement
+
+  groupVisibleElement = [
+    'visibleElements',
+    'visibleAnchors',
+    'visiblePPRE',
+    'visibleAnchorOverPPre',
+  ]
+  groups['visibleElement'] = groupVisibleElement
+
+  groupEntries = [
+    'numSection',
+    'numSection2',
+    'numSection3',
+    'numArticle',
+    'numArticle2',
+    'numArticle3',
+    'numEntries',
+    'numEntries2',
+    'numEntries3',
+    'numH1',
+    'numH2',
+    'numH3',
+    'numH4',
+    'headCountSum',
+    'headCountMax',
+    'entryCountSum',
+    'entryCountMax',
+  ]
+  groups['entries'] = groupEntries
+
+  groupV1 = [
+    'openGraph',
+
+    'forum',
+    'index',
+    'search',
+    'view',
+    'archive',
+    'asp',
+    'phpbb',
+    'php',
+    'pathLength',
+    'domain',
+    'pathComponents',
+    'slugDetector',
+    'pathNumbers',
+    'lastSegmentLength',
+
+    'formCount',
+    'anchorCount',
+    'elementCount',
+    'anchorRatio',
+
+    'mozScore',
+    'mozScoreAllSqrt',
+    'mozScoreAllLinear',
+  ]
+  groups['v1'] = groupV1
+
+  groupV1NoPath = [
+    'openGraph',
+
+    'formCount',
+    'anchorCount',
+    'elementCount',
+    'anchorRatio',
+
+    'mozScore',
+    'mozScoreAllSqrt',
+    'mozScoreAllLinear',
+  ]
+  groups['v1NoPath'] = groupV1NoPath
+
+  groups['allElement'] = groupNumElement + groupVisibleElement + groupEntries
+  groups['mozScores'] = [f for f in header if 'moz' in f]
+  groups['noText'] = [f for f in header if not ('inner' in f or 'Content' in f or 'WordCount' in f)]
+
+  return groups
 
 def main(argv):
   parser = argparse.ArgumentParser()
-  parser.add_argument('--out', required=True)
-  parser.add_argument('--marked', required=True)
-  parser.add_argument('--features', required=True)
+  parser.add_argument('--out', required=True, help="filename of output")
+  parser.add_argument('--marked', help="filename of marked output")
+  parser.add_argument('--distilled', help="filename of derived features of distilled content")
+  parser.add_argument('--features', required=True, help="filename of derived features")
   options = parser.parse_args(argv)
 
-  marked = None
-  with open(options.marked) as markedin:
-    marked = json.load(markedin)
-
-  features = None
+  if (options.marked is None) + (options.distilled is None) != 1:
+    print 'Use exactly one of --marked or --distilled.'
+    os.exit(1)
+
   with open(options.features) as features:
     features = json.load(features)
 
-  markedMap = dict()
-  # good:
-  # -1 error
-  # 0 bad
-  # 1 good
-  # 2 good w/error
-  for m in marked:
-    if not 'good' in m:
-      continue
-    if m['good'] < 0:
-      continue
-    markedMap[m['url']] = m
-
-  merged = []
-  for f in features:
-    url = f['url']
-    if not url in markedMap:
-      continue
-    merged.append(map(float, [0 if markedMap[url]['good'] == 0 else 1] + f['features'][1::2]))
-
-  header = ['good'] + map(str, features[0]['features'][::2])
-
-  with open(options.out, 'w') as csvfile:
-    writer = csv.writer(csvfile)
-    writer.writerow(header)
-    for e in merged:
-      writer.writerow(e)
+  if options.marked:
+    with open(options.marked) as markedin:
+      marked = json.load(markedin)
+
+    markedMap = dict()
+    # good:
+    # -1 error
+    # 0 bad
+    # 1 good
+    # 2 good w/error
+    for m in marked:
+      if not 'good' in m:
+        continue
+      if m['good'] < 0:
+        continue
+      markedMap[m['url'].strip()] = m
+
+    print "Loaded %d labeled entries" % (len(markedMap))
+
+    merged = []
+    for f in features:
+      url = f['url']
+      if not url in markedMap:
+        continue
+      merged.append(map(float, [0 if markedMap[url]['good'] == 0 else 1] + f['features'][1::2]))
+    print "Merged %d entries" % (len(merged))
+
+  if options.distilled:
+    with open(options.distilled) as markedin:
+      marked = json.load(markedin)
+
+    markedMap = dict()
+    for m in marked:
+      feature = m['features']
+      feature = dict(zip(feature[::2], feature[1::2]))
+      if feature['innerTextLength'] == 0:
+        continue
+      m['features'] = feature
+      markedMap[m['url'].strip()] = m
+
+    print "Loaded %d distilled entries" % (len(markedMap))
+
+    merged = []
+    for f in features:
+      url = f['url']
+      if not url in markedMap:
+        continue
+      if f['native']['features']['isMobileFriendly'] == 1:
+        continue
+      if f['native']['distillable'] != 1:
+        continue
+      feature = markedMap[url]['features']
+      merged.append(map(float, [0 if feature['innerTextLength'] < 1000 else 1] + f['features'][1::2]))
+    print "Merged %d entries" % (len(merged))
+
+  feature_headers = map(str, features[0]['features'][::2])
+  header = ['good'] + feature_headers
+
+  write_features(options.out, header, merged, header)
+
+  # write datasets with a single feature
+  outbase = os.path.splitext(options.out)[0]
+  for s in feature_headers:
+    print 'Single feature: %s' % s
+    write_features('%s-feature-%s.csv' % (outbase, s), header, merged, ['good', s])
+
+  # write datasets with feature groups
+  for (name, g) in getGroups(feature_headers).iteritems():
+    print 'Feature group: %s' % name
+    write_features('%s-group-%s.csv' % (outbase, name), header, merged, ['good'] + g)
 
   return 0
 
 if __name__ == '__main__':
   sys.exit(main(sys.argv[1:]))
-