def parse_zipped_j3m(uv_task):
task_tag = "PARSING ZIPPED J3M"
print "\n\n************** %s [START] ******************\n" % task_tag
print "parsing zipped j3m asset at %s" % uv_task.doc_id
uv_task.setStatus(302)
import os
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
media = UnveillanceDocument(_id=uv_task.doc_id)
if media is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
from conf import ANNEX_DIR
if hasattr(uv_task, "j3m_name"):
j3m_name = uv_task.j3m_name
else:
j3m_name = os.path.join(media.base_path, "j3m_raw.gz")
if not media.getFile(j3m_name):
print "NO J3M.GZ at %s" % j3m_name
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
from cStringIO import StringIO
from lib.Worker.Utils.funcs import getFileType, unGzipBinary
from vars import MIME_TYPES
j3m = media.loadFile(j3m_name)
j3m_type = getFileType(j3m, as_buffer=True)
if j3m_type == MIME_TYPES['gzip']:
j3m = unGzipBinary(j3m)
if j3m is None or getFileType(j3m, as_buffer=True) != MIME_TYPES['json']:
print "THIS IS NOT A J3M (type %s)" % j3m_type
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail(status=412)
return
asset_path = "j3m_raw.json"
media.addAsset(j3m, asset_path, as_literal=False)
uv_task.put_next([
"J3M.j3mify.j3mify",
"J3M.massage_j3m.massageJ3M",
"PGP.verify_signature.verifySignature",
"J3M.verify_visual_content.verifyVisualContent"
])
uv_task.routeNext(inflate={'j3m_name' : asset_path})
uv_task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def preprocessNLP(task):
task_tag = "TEXT NLP PREPROCESSING"
print "\n\n************** %s [START] ******************\n" % task_tag
print "nlp preprocessing text at %s" % task.doc_id
task.setStatus(302)
import re
from json import loads
from lib.Worker.Models.uv_document import UnveillanceDocument
from lib.Core.Utils.funcs import cleanAndSplitLine
from conf import DEBUG
from vars import ASSET_TAGS
document = UnveillanceDocument(_id=task.doc_id)
if document is None:
print "DOC IS NONE"
task.fail()
return
# 1. get all the words (bag of words)
try:
texts = loads(document.loadAsset("doc_texts.json"))
except Exception as e:
print "ERROR GETTING DOC-TEXTS: %s" % e
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
word_groups = [cleanAndSplitLine(text) for text in texts if text is not None]
word_groups = [wg for wg in word_groups if len(wg) > 0]
bag_of_words = sum(word_groups, [])
document.addAsset(bag_of_words, "bag_of_words.txt", as_literal=False,
description="bag of words", tags=ASSET_TAGS['BOW'])
# 2. get keywords, weighted and parsable by gensim
once_words = set(word for word in set(bag_of_words) if bag_of_words.count(word) == 1)
key_words = [word for word in bag_of_words if word not in once_words]
if len(key_words) > 0:
document.addAsset(key_words, "key_words_gensim.txt", as_literal=False,
description="keywords, as list, and parsable by gensim",
tags=ASSET_TAGS['KW'])
document.addCompletedTask(task.task_path)
task.routeNext()
task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def extractNEREntities(task):
task_tag = "NER ENTITY EXTRACTION"
print "\n\n************** %s [START] ******************\n" % task_tag
print "TOKENIZING TEXT DOCUMENT at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
doc = UnveillanceDocument(_id=task.doc_id)
if doc is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
from json import loads
try:
texts = loads(doc.loadAsset("doc_texts.json"))
except Exception as e:
print "ERROR GETTING DOC-TEXTS: %s" % e
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
import ner, os
from conf import getConfig
from lib.Core.Utils.funcs import cleanLine
st = ner.SocketNER(host='localhost', port=getConfig("nlp_server.port"))
entities = {}
for i, page in enumerate(texts):
if page is None: continue
lemmas = st.get_entities(cleanLine(page))
if len(lemmas.keys()) == 0: continue
for lemma_type in lemmas.keys():
entities = updateEntities(entities, lemmas[lemma_type], lemma_type, i)
#if DEBUG and i > 25: break
if len(entities.keys()) > 0:
ner_entity_path = doc.addAsset(entities, "stanford-ner_entities.json", as_literal=False,
description="Entities as per Stanford-NER Tagger (via NLTK)",
tags=[ASSET_TAGS['STANFORD_NER_ENTITIES'], ASSET_TAGS['CP_ENTITIES']])
if ner_entity_path is not None:
doc.addFile(ner_entity_path, None, sync=True)
doc.addCompletedTask(task.task_path)
task.routeNext()
print "\n\n************** %s [END] ******************\n" % task_tag
task.finish()
def get_vector(uv_task): task_tag = "IMAGE: GETTING VECTOR" print "\n\n************** %s [START] ******************\n" % task_tag uv_task.setStatus(302) from lib.Worker.Models.uv_document import UnveillanceDocument from vars import ASSET_TAGS from conf import ANNEX_DIR, DEBUG import os, pypuzzle image = UnveillanceDocument(_id=uv_task.doc_id) hi_res = image.getAssetsByTagName(ASSET_TAGS['HIGH']) if hi_res is None: error_msg = "Could not find the hi-res clone" print error_msg print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_msg) return hi_res = os.path.join(ANNEX_DIR, image.base_path, hi_res[0]['file_name']) puzz = pypuzzle.Puzzle() if DEBUG: print "generate puzzle vector from %s" % hi_res try: cvec = puzz.get_cvec_from_file(hi_res) except Exception as e: error_msg = "Could not get image vector because %s" % e print error_msg print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_msg) return if not image.addAsset(cvec, "image_cvec.json", as_literal=False, tags=[ASSET_TAGS['IMAGE_CVEC']]): error_msg = "could not save cvec asset!" print error_msg print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_msg) return print "\n\n************** %s [END] ******************\n" % task_tag uv_task.finish()
def get_image_vector(uv_task):
task_tag = "AVI: GETTING IMAGE VECTOR"
print "\n\n************** %s [START] ******************\n" % task_tag
uv_task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import ANNEX_DIR
import pypuzzle
image = UnveillanceDocument(_id=uv_task.doc_id)
puzz = pypuzzle.Puzzle()
try:
cvec = puzz.get_cvec_from_file(os.path.join(ANNEX_DIR,
image.file_name))
except Exception as e:
error_msg = "Could not get image vector because %s" % e
print error_msg
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail(message=error_msg)
return
from vars import ASSET_TAGS
if not image.addAsset(cvec,
"image_cvec.json",
as_literal=False,
tags=[ASSET_TAGS['IMAGE_CVEC']]):
error_msg = "could not save cvec asset!"
print error_msg
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail(message=error_msg)
return
print "\n\n************** %s [END] ******************\n" % task_tag
uv_task.finish()
def get_image_vector(uv_task): task_tag = "AVI: GETTING IMAGE VECTOR" print "\n\n************** %s [START] ******************\n" % task_tag uv_task.setStatus(302) from lib.Worker.Models.uv_document import UnveillanceDocument from conf import ANNEX_DIR import pypuzzle image = UnveillanceDocument(_id=uv_task.doc_id) puzz = pypuzzle.Puzzle() try: cvec = puzz.get_cvec_from_file(os.path.join(ANNEX_DIR, image.file_name)) except Exception as e: error_msg = "Could not get image vector because %s" % e print error_msg print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_msg) return from vars import ASSET_TAGS if not image.addAsset(cvec, "image_cvec.json", as_literal=False, tags=[ASSET_TAGS['IMAGE_CVEC']]): error_msg = "could not save cvec asset!" print error_msg print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_msg) return print "\n\n************** %s [END] ******************\n" % task_tag uv_task.finish()
def evaluateText(task):
task_tag = "TEXT EVALUATION"
print "\n\n************** %s [START] ******************\n" % task_tag
print "evaluating text at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import MIME_TYPE_TASKS
document = UnveillanceDocument(_id=task.doc_id)
"""
limited choices: json, pgp, or txt
"""
if hasattr(task, "text_file"):
content = document.loadAsset(task.text_file)
else:
content = document.loadFile(document.file_name)
if content is None:
print "no text to evaluate :("
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
new_mime_type = None
import json
try:
json_txt = json.loads(content)
new_mime_type = "application/json"
print "THIS IS JSON"
except Exception as e:
print "NOT JSON: %s" % e
task_path = None
if new_mime_type is not None:
document.mime_type = new_mime_type
document.save()
if document.mime_type in MIME_TYPE_TASKS.keys():
task_path = MIME_TYPE_TASKS[document.mime_type][0]
else:
try:
from lib.Core.Utils.funcs import cleanLine
from vars import ASSET_TAGS
txt_json = []
txt_pages = []
line_count = 0
# this is arbitrary
MAX_LINES_PER_PAGE = 80
for line in content.splitlines():
txt_pages.append(cleanLine(line))
line_count += 1
if line_count == MAX_LINES_PER_PAGE:
txt_json.append(" ".join(txt_pages))
txt_pages = []
line_count = 0
txt_json.append(" ".join(txt_pages))
document.total_pages = len(txt_json)
document.save()
asset_path = document.addAsset(txt_json, "doc_texts.json", as_literal=False,
description="jsonified text of original document, segment by segment",
tags=[ASSET_TAGS['TXT_JSON']])
from lib.Worker.Models.uv_text import UnveillanceText
uv_text = UnveillanceText(inflate={
'media_id' : document._id,
'searchable_text' : txt_json,
'file_name' : asset_path
})
document.text_id = uv_text._id
document.save()
except Exception as e:
if DEBUG:
print "ERROR HERE GENERATING DOC TEXTS:"
print e
document.addCompletedTask(task.task_path)
task.finish()
task.routeNext()
print "\n\n************** %s [END] ******************\n" % task_tag
def massageJ3M(task):
task_tag = "MASSAGING J3M"
print "\n\n************** %s [START] ******************\n" % task_tag
print "massaging j3m at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
media = UnveillanceDocument(_id=task.doc_id)
if media is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
if hasattr(task, "j3m_name"):
j3m_name = task.j3m_name
else:
j3m_name = "j3m.json"
j3m = media.loadAsset(j3m_name)
if j3m is None:
print "J3M IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
from json import loads
try:
j3m = loads(j3m)
except Exception as e:
print "J3M IS INVALID"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail(status=412)
return
try:
media.date_created = j3m['genealogy']['dateCreated']
media.saveFields("date_created")
except KeyError as e:
print "J3M HAS NO DATE CREATED: %s" % e
print "\n\n************** %s [WARN] ******************\n" % task_tag
from hashlib import sha1
try:
j3m['public_hash'] = sha1("".join(
[j3m['genealogy']['createdOnDevice'],
"".join(j3m['genealogy']['hashes'])])).hexdigest()
except KeyError as e:
if DEBUG:
print "no key %s" % e
pass
if 'data' in j3m.keys():
try:
location = j3m['data']['exif']['location']
j3m['data']['exif'].update({
'location' : [location[1], location[0]]
})
except KeyError as e:
if DEBUG:
print "no key %s" % e
pass
try:
if type(j3m['data']['sensorCapture']) is list: pass
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
if 'sensorCapture' in j3m['data'].keys():
for playback in j3m['data']['sensorCapture']:
if 'gps_coords' in playback['sensorPlayback'].keys():
try:
gps = str(playback['sensorPlayback']['gps_coords'])[1:-1].split(",")
if DEBUG:
print "REPLACING %s as geopoint" % gps
print type(gps)
playback['sensorPlayback'].update({
'gps_coords' : [float(gps[1]), float(gps[0])]
})
except Exception as e:
if DEBUG: print e
pass
if 'regionLocationData' in playback['sensorPlayback'].keys():
try:
gps = str(playback['sensorPlayback']['regionLocationData']['gps_coords'])
gps = gps[1:-1].split(",")
if DEBUG:
print "REPLACING %s as geopoint" % gps
playback['sensorPlayback']['regionLocationData'].update({
'gps_coords' : [float(gps[1]), float(gps[0])]
})
except Exception as e:
if DEBUG: print e
pass
if 'visibleWifiNetworks' in playback['sensorPlayback'].keys():
try:
for i,b in enumerate(playback['sensorPlayback']['visibleWifiNetworks']):
playback['sensorPlayback']['visibleWifiNetworks'][i].update({
'bt_hash' : sha1(b['bssid']).hexdigest()
})
except Exception as e:
if DEBUG: print e
pass
import os, json
from conf import getConfig
from lib.Core.Utils.funcs import b64decode
from lib.Worker.Utils.funcs import getFileType, unGzipBinary
searchable_text = []
if 'userAppendedData' in j3m['data'].keys():
try:
with open(os.path.join(getConfig('informacam.forms_root'), "forms.json"), 'rb') as F:
form_data = json.loads(F.read())['forms']
for udata in j3m['data']['userAppendedData']:
for aForms in udata['associatedForms']:
st_keys = aForms['answerData'].keys()
for f in form_data:
if f['namespace'] == aForms['namespace']:
try:
for mapping in f['mapping']:
try:
group = mapping.keys()[0]
key = aForms['answerData'][group].split(" ")
for m in mapping[group]:
if m.keys()[0] in key:
key[key.index(m.keys()[0])] = m[m.keys()[0]]
aForms['answerData'][group] = " ".join(key)
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
try:
idx = 0
for audio in f['audio_form_data']:
try:
while audio in st_keys: st_keys.remove(audio)
except Exception as e: pass
try:
audio_data = b64decode(
aForms['answerData'][audio])
if audio_data is None:
if DEBUG: print "could not unb64 audio"
continue
if getFileType(audio_data, as_buffer=True) != MIME_TYPES['gzip']:
if DEBUG: print "audio is not gzipped"
continue
audio_f = "audio_%d.3gp" % idx
idx += 1
media.addAsset(unGzipBinary(audio_data), audio_f,
tags=[ASSET_TAGS['A_3GP']],
description="3gp audio file from form")
'''
new_task=UnveillanceTask(inflate={
'task_path' : "Media.convert.audioConvert",
'doc_id' : media._id,
'formats' : ["3gp", "wav"],
'src_file' : "audio_%d.3gp" % idx,
'queue' : task.queue
})
new_task.run()
'''
aForms['answerData'][audio] = "audio_%d.wav"
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
if len(st_keys) > 0:
for key in st_keys:
searchable_text.append(aForms['answerData'][key])
except KeyError as e:
if DEBUG: print "no key %s" % e
pass
except IOError as e:
print "\n\n************** %s [WARN] ******************\n" % task_tag
if DEBUG: print "no forms to go over: %s" % e
except ValueError as e:
print "\n\n************** %s [WARN] ******************\n" % task_tag
if DEBUG: print "for some reason, forms.json is not legible?\n%s" % e
if media.addAsset(j3m, "j3m.json", as_literal=False) is False:
print "J3M COULD NOT BE ADDED"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
from lib.Worker.Models.ic_j3m import InformaCamJ3M
j3m['media_id'] = media._id
if len(searchable_text) > 0:
j3m['searchable_text'] = searchable_text
j3m = InformaCamJ3M(inflate=j3m)
print "\n\n***NEW J3M CREATED***\n\n"
j3m.save()
media.j3m_id = j3m._id
media.save()
media.addCompletedTask(task.task_path)
task.routeNext()
task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def locate_j3m(uv_task):
task_tag = "PULLING J3M"
print "\n\n************** %s [START] ******************\n" % task_tag
print "pulling j3m at %s" % uv_task.doc_id
uv_task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG, ANNEX_DIR
from vars import ASSET_TAGS
media = UnveillanceDocument(_id=uv_task.doc_id)
if media is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
from lib.Worker.Utils.funcs import getFileType
from vars import MIME_TYPES, MIME_TYPE_MAP
ic_j3m_txt = media.loadAsset("j3m_raw.txt")
ic_j3m_txt_mime_type = getFileType(ic_j3m_txt, as_buffer=True, force_json=True)
inflate = {}
print "J3M MIME TYPE SNIFFED: %s" % ic_j3m_txt_mime_type
if ic_j3m_txt_mime_type != MIME_TYPES['json']:
import os
from lib.Core.Utils.funcs import b64decode
un_b64 = b64decode(ic_j3m_txt)
if un_b64 is not None:
un_b64_mime_type = getFileType(un_b64, as_buffer=True)
if un_b64_mime_type in [MIME_TYPES['pgp'], MIME_TYPES['gzip']]:
if DEBUG:
print "MIME TYPE: %s" % un_b64_mime_type
asset_path = "j3m_raw.%s" % MIME_TYPE_MAP[un_b64_mime_type]
media.addAsset(un_b64, asset_path)
if DEBUG:
print "\n\nPGP KEY FILE PATH: %s\n\n" % asset_path
gz = media.addAsset(None, "j3m_raw.gz", tags=[ASSET_TAGS['OB_M']],
description="j3m data extracted from obscura marker")
if un_b64_mime_type == MIME_TYPES['pgp']:
uv_task.put_next([
"PGP.decrypt.decrypt",
"J3M.j3mify.parse_zipped_j3m"
])
inflate.update({
'pgp_file' : os.path.join(media.base_path, asset_path),
'save_as' : gz
})
was_encrypted = True
elif un_b64_mime_type in MIME_TYPES['gzip']:
uv_task.put_next("J3M.j3mify.parse_zipped_j3m")
else:
import os
from fabric.api import settings, local
with settings(warn_only=True):
src_j3m = os.path.join(ANNEX_DIR, media.base_path, "j3m_raw.txt")
dest_j3m = os.path.join(ANNEX_DIR, media.base_path, "j3m_raw.json")
local("mv %s %s" % (src_j3m, dest_j3m))
print "PUTTING J3M FROM HERE!!!! WAS JSON! (%s -> %s)" % (src_j3m, dest_j3m)
media.addAsset(None, "j3m_raw.json")
uv_task.put_next([
"J3M.j3mify.j3mify",
"PGP.verify_signature.verifySignature",
"J3M.massage_j3m.massageJ3M",
"J3M.verify_visual_content.verifyVisualContent",
"J3M.notarize.notarize_media"
])
inflate.update({'j3m_name' : "j3m_raw.json"})
media.addCompletedTask(uv_task.task_path)
uv_task.routeNext(inflate=inflate)
uv_task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def basicTokenizer(task):
task_tag = "NLP TOKENIZER"
print "\n\n************** %s [START] ******************\n" % task_tag
print "TOKENIZING TEXT DOCUMENT at %s" % task.doc_id
task.setStatus(412)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
doc = UnveillanceDocument(_id=task.doc_id)
if doc is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
return
txt = None
from json import loads
if hasattr(task, "txt_file"):
txt = loads(doc.loadFile(task.txt_file))
else:
import os
try:
txt_path = doc.getAssetsByTagName(ASSET_TAGS['TXT_JSON'])[0]['file_name']
txt = loads(doc.loadFile(os.path.join(doc.base_path, txt_path)))
except Exception as e:
if DEBUG: print e
if txt is None:
print "TEXT FILE IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
return
from lib.Worker.Models.cp_nlp_server import CompassNLPServer
nlp_server = CompassNLPServer()
tokenized = nlp_server.sendNLPRequest({
'method' : 'tokenize',
'txt' : txt
})
if tokenized is None:
print "COULD NOT TOKENIZE."
print "\n\n************** %s [ERROR] ******************\n" % task_tag
return
if DEBUG:
print "here is res"
print type(tokenized)
asset_path = doc.addAsset(tokenized, "core_nlp_tokenized.json", as_literal=False,
description="tokenized output from Stanford Core NLP",
tags=[ASSET_TAGS['TOKENS_NLP']])
if asset_path is None or not doc.addFile(asset_path, None, sync=True):
print "COULD NOT SAVE ASSET."
print "\n\n************** %s [ERROR] ******************\n" % task_tag
return
doc.addCompletedTask(task.task_path)
task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def compileMetadata(task):
task_tag = "COMPILING METADATA"
print "\n\n************** %s [START] ******************\n" % task_tag
print "compiling metadata for %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
document = UnveillanceDocument(_id=task.doc_id)
if document is None:
err = "DOC IS NONE"
print err
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail(message=err)
return
metadata = document.loadAsset(task.md_file)
if metadata is None:
print "NO METADATA FILE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
return
import csv, re
from Levenshtein import ratio
from string import letters
from vars import METADATA_ASPECTS, ASSET_TAGS
numbers = str("".join([str(i) for i in range(0,10)]))
missing_value = "NaN"
labels = ["_id"]
values = [document._id]
try:
for mda in METADATA_ASPECTS[task.md_namespace]:
labels.append(mda['label'])
if hasattr(task, "md_rx"):
pattern = re.compile(task.md_rx % (mda['tag_position'], mda['label']))
else:
pattern = re.compile(mda['tag_position'])
if DEBUG: print pattern.pattern
value = missing_value
ideal = mda['ideal']
if mda['ideal'] is None:
if mda['type'] == "str":
ideal = letters + numbers
elif mda['type'] == "int":
ideal = int(numbers)
print "IDEAL FOR TAG: %s" % ideal
for line in metadata.splitlines():
match = re.findall(pattern, line.strip())
if len(match) == 1:
if DEBUG: print "VALUE FOUND: %s (%s)" % (match[0], type(match[0]))
if mda['type'] == "str":
try:
value = "%.9f" % ratio(ideal, str(value.replace("\"", '')))
except TypeError as e:
if DEBUG: print e
value = 0
elif mda['type'] == "int":
try:
value = ideal/float(match[0].replace("\"", ''))
except ZeroDivisionError as e:
if DEBUG: print e
value = 0
break
if value == missing_value:
if mda['ideal'] is None: value = 1
else: value = 0
values.append(value)
if hasattr(task, 'md_extras'):
for key, value in task.md_extras.iteritems():
labels.append(key)
values.append(value)
"""
if DEBUG:
print "labels %s" % labels
print "values %s" % values
"""
from cStringIO import StringIO
md_csv_file = StringIO()
md_csv = csv.writer(md_csv_file,
delimiter=',', quotechar='|', quoting=csv.QUOTE_MINIMAL)
md_csv.writerow(labels)
md_csv.writerow(values)
md_asset = document.addAsset(md_csv_file.getvalue(), "file_metadata.csv",
tags=[ASSET_TAGS["F_MD"]],
description="CSV representation of %s" % task.md_file)
md_csv_file.close()
if md_asset is None or not document.addFile(md_asset, None):
print "Could not save the Metadata"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
document.addCompletedTask(task.task_path)
from lib.Worker.Utils.funcs import routeNextTask
routeNextTask(task, document)
task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
except KeyError as e:
if DEBUG: print e
print "No metadata aspects for %s" % task.md_namespace
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
def addressParser(task):
task_tag = "NLP ADDRESS PARSER"
print "\n\n************** %s [START] ******************\n" % task_tag
print "EXTRACTING ADDRESSES FROM TEXT DOCUMENT at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
doc = UnveillanceDocument(_id=task.doc_id)
if doc is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
txt = None
if hasattr(task, "txt_file"):
txt = doc.loadFile(task.txt_file)
else:
import os
try:
txt_path = doc.getAssetsByTagName(ASSET_TAGS["TXT_JSON"])[0]["file_name"]
txt = doc.loadFile(os.path.join(doc.base_path, txt_path))
except Exception as e:
if DEBUG:
print e
if txt is None:
print "TEXT FILE IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
import re
# script from https://code.google.com/p/ebcode/ -> ebdata.tar.gz -> ebdata/nlp/addresses.py
# Regex notes:
# * This is *not* a case-insensitive regex, because we assume
# capitalized words are special (street names).
# * All data matched by capturing parentheses is concatenated together, so
# if you don't want to include something in the resulting string, don't
# capture it.
# STREET_NAME is a fragment of a regular expression that is used in several
# places in our "real" regular expression (ADDRESSES_RE) below. The one tricky
# thing about it is that it includes a "CAPTURE_START" placeholder instead of
# a capturing opening parenthesis. This lets us create two versions of the
# regex -- STREET_NAME_CAPTURE and STREET_NAME_NOCAPTURE.
STREET_NAME = r"""
# Here, we define some common false positives and tell the regex to ignore them.
(?!
[Aa][Ss][Ss][Oo][Cc][Ii][Aa][Tt][Ee][Dd]\ [Pp][Rr][Ee][Ss][Ss] # associated press
|
[Uu][Nn][Ii][Vv][Ee][Rr][Ss][Ii][Tt][Yy]\ [Oo][Ff] # university of
)
# DIRECTION
%(CAPTURE_START)s
(?:
[NSEWnsew]\.?
|
(?:
[Nn][Oo][Rr][Tt][Hh] |
[Ss][Oo][Uu][Tt][Hh] |
[Ee][Aa][Ss][Tt] |
[Ww][Ee][Ss][Tt] |
[Nn][Oo][Rr][Tt][Hh][Ee][Aa][Ss][Tt] |
[Ee][Aa][Ss][Tt][Ww][Ee][Ss][Tt] |
[Ss][Oo][Uu][Tt][Hh][Ee][Aa][Ss][Tt] |
[Ss][Oo][Uu][Tt][Hh][Ww][Ee][Ss][Tt]
)
|
(?:
N\.?W | S\.?W | N\.?E | S\.?E
)\.?
)
\ + # space (but not newline)
)?
(?:
# STREET NAME
%(CAPTURE_START)s
# Numbered street names with a suffix ("3rd", "4th").
\d+(?:st|ST|nd|ND|rd|RD|th|TH|d|D)
|
# Or, numbered street names without a suffix ("3", "4")
# but with a street type.
\d+
(?=
\ +
(?:Ave|Avenue|Blvd|Boulevard|Bvd|Cir|Circle|Court|Ct|Dr|Drive|
Lane|Ln|Parkway|Pkwy|Place|Plaza|Pl|Plz|Point|Pt|Pts|Rd|Rte|
Sq|Sqs|Street|Streets|St|Sts|Terrace|Ter|Terr|Trl|Way|Wy
)
\b
)
|
# Or, street names that don't start with numbers.
(?:
# Optional prefixes --
# "St", as in "St Louis"
# "Dr. Martin", as in "Dr. Martin Luther King"
(?:
[Ss][Tt]\.?
|
[Dd][Rr]\.?\ [Mm][Aa][Rr][Tt][Ii][Nn]
)
\ +
)?
(?:
Mass\.(?=\ +[Aa]ve) # Special case: "Mass." abbr. for "Massachussetts Ave."
# Needs to be special-cased because of the period.
|
(?:Avenue|Ave\.?)\ +[A-Z] # Special case: "Avenue X"
|
[A-Z][a-z][A-Za-z]* # One initial-capped word
|
[A-Z]\b # Single-letter street name (e.g., K St. in DC)
(?!\.\w) # Avoid '20 U.S.A.'
)
)
(?:
# Here, we list the options with street suffixes first, so that
# the suffix abbreviations are treated as the last part of the
# street name, to avoid overeagerly capturing "123 Main St. The".
%(CAPTURE_START)s
\ +(?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
\ +[A-Z][a-z][A-Za-z]*\ (?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
(?:,?\ Jr\.?,?|\ +[A-Z][a-z][A-Za-z]*){2}\ +(?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
(?:,?\ Jr\.?,?|\ +[A-Z][a-z][A-Za-z]*){3}\ +(?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
(?:,?\ Jr\.?,?|\ +[A-Z][a-z][A-Za-z]*){4}\ +(?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
(?:,?\ Jr\.?,?|\ +[A-Z][a-z][A-Za-z]*){5}\ +(?:Ave|Blvd|Bvd|Cir|Ct|Dr|Ln|Pkwy|Pl|Plz|Pt|Pts|Rd|Rte|Sq|Sqs|St|Sts|Ter|Terr|Trl|Wy)\.
|
(?:,?\ Jr\.?,?|\ +[A-Z][a-z][A-Za-z]*){1,5}
)?
# OPTIONAL POST-DIR
(?:
# Standard post-dir format
%(CAPTURE_START)s
,?\s(?:N\.?E|S\.?E|N\.?W|S\.?W|N|S|E|W)\.?
)
# Avoid greedily capturing more letters, like
# '123 Main St, New England' to '123 Main St, N'
(?![A-Za-z])
|
# Or, a special-case for DC quadrants, to find stuff like:
# "600 H Street in NE Washington"
# "600 H Street in the NE quadrant"
# "600 H Street in northeast DC"
# Note that this is NOT captured, so that it's excluded from
# the final output.
,?
\s in
%(CAPTURE_START)s
\s
)
(?:
(?:the|far) \s
)?
%(CAPTURE_START)s
(?:NE|SE|NW|SW|[Nn]ortheast|[Ss]outheast|[Nn]orthwest|[Ss]outhwest)
(?=
\s (?:quadrant|D\.?C\.?|Washington)
)
)
)?
)?
)
"""
STREET_NAME_CAPTURE = STREET_NAME % {"CAPTURE_START": "("}
STREET_NAME_NOCAPTURE = STREET_NAME % {"CAPTURE_START": "(?:"}
ADDRESSES_RE = re.compile(
r"""(?x)
(?<!-|/|:|,|\.|\$) # These various characters are not allowed before an address/intersection.
\b
# Ignore things that look like dates -- e.g., "21 May 2009".
# This is a problem e.g. in cases where there's a May Street.
(?!
\d+\s+
(?:January|February|March|April|May|June|July|August|September|October|November|December)
,?\s+
\d\d\d\d
)
# Ignore intersections that are prefixed by "University of", like
# "University of Texas at Austin". This is a common false positive.
(?<!
[Uu][Nn][Ii][Vv][Ee][Rr][Ss][Ii][Tt][Yy]\s[Oo][Ff]\s
)
(?:
# SEGMENT ("FOO BETWEEN BAR AND BAZ")
(?:
%(STREET_NAME_CAPTURE)s (,?\ + between \ +) %(STREET_NAME_CAPTURE)s (,?\ + and \ +) %(STREET_NAME_CAPTURE)s
|
%(STREET_NAME_CAPTURE)s (,?\ + from \ +) %(STREET_NAME_CAPTURE)s (,?\ + to \ +) %(STREET_NAME_CAPTURE)s
)
|
# BLOCK/ADDRESS
(?:
(
(?:
(?:\d+|[Ff][Ii][Rr][Ss][Tt])[-\ ]
(?:(?:[Nn][Oo][Rr][Tt][Hh]|[Ss][Oo][Uu][Tt][Hh]|[Ee][Aa][Ss][Tt]|[Ww][Ee][Ss][Tt])\ )?
[Bb][Ll][Oo][Cc][Kk]\ [Oo][Ff]
|
\d+\ *-\ *\d+
|
\d+
)
\ +
)
%(STREET_NAME_CAPTURE)s
# ignore the intersection in parenthesis so that it's not picked
# up as a separate location. We do this by consuming the string
# but *not* capturing it.
(?:
\ +
\(?
between
\ +
%(STREET_NAME_NOCAPTURE)s
\ +
and
\ +
%(STREET_NAME_NOCAPTURE)s
\)?
)?
)
|
# INTERSECTION
(?:
# Common intersection prefixes. They're included here so that the
# regex doesn't include them as part of the street name.
(?:
(?:
[Nn]ear |
[Aa]t |
[Oo]n |
[Tt]o |
[Aa]round |
[Ii]ntersection\ of |
[Cc]orner\ of |
[Aa]rea\ of |
[Aa]reas?\ surrounding |
vicinity\ of |
ran\ down |
running\ down |
crossed
)
\ +
)?
\b
(?:%(STREET_NAME_CAPTURE)s)
(\ +)
(
(?:
[Aa][Nn][Dd] |
[Aa][Tt] |
[Nn][Ee][Aa][Rr] |
& |
[Aa][Rr][Oo][Uu][Nn][Dd] |
[Tt][Oo][Ww][Aa][Rr][Dd][Ss]? |
[Oo][Ff][Ff] |
(?:[Jj][Uu][Ss][Tt]\ )?(?:[Nn][Oo][Rr][Tt][Hh]|[Ss][Oo][Uu][Tt][Hh]|[Ee][Aa][Ss][Tt]|[Ww][Ee][Ss][Tt])\ [Oo][Ff] |
(?:[Jj][Uu][Ss][Tt]\ )?[Pp][Aa][Ss][Tt]
)
\ +
)
(?:%(STREET_NAME_CAPTURE)s)
)
)
# OPTIONAL CITY SUFFIX
(?:
(?:
,?\s+in |
,
)
\s+
# CITY NAME
(
[A-Z][a-z][A-Za-z]* # One initial-capped word
(?:
,?\ Jr\.?,?
|
\ [A-Z][a-z][A-Za-z]*
|
-[A-Za-z]+ # Hyphenated words (e.g. "Croton-on-Hudson" in NY)
){0,4} # Initial-capped words
)
)?
"""
% {"STREET_NAME_CAPTURE": STREET_NAME_CAPTURE, "STREET_NAME_NOCAPTURE": STREET_NAME_NOCAPTURE}
)
addresses = parse_addresses(txt, ADDRESSES_RE)
if addresses is None:
print "COULD NOT EXTRACT ADDRESSES."
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
asset_path = doc.addAsset(
addresses,
"addresses.json",
as_literal=False,
description="addresses output from Everyblock address extractor",
tags=[ASSET_TAGS["ADDRESSES_NLP"], ASSET_TAGS["CP_ENTITIES"]],
)
if asset_path is None or not doc.addFile(asset_path, None, sync=True):
print "COULD NOT SAVE ASSET."
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
doc.addCompletedTask(task.task_path)
task.routeNext()
print "\n\n************** %s [END] ******************\n" % task_tag
task.finish()
def createGensimObjects(task):
task_tag = "GENSIM TOPIC EXTRACTION"
print "\n\n************** %s [START] ******************\n" % task_tag
print "USING TEXT DOCUMENT at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
doc = UnveillanceDocument(_id=task.doc_id)
if doc is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
from json import loads
try:
texts = loads(doc.loadAsset("doc_texts.json"))
except Exception as e:
print "ERROR GETTING DOC-TEXTS: %s" % e
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
if len(texts) == 0:
print "THERE ARE NO TEXTS HERE ANYWAY!"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
import logging, os, bz2
from json import loads
from gensim import corpora
from lib.Core.Utils.funcs import cleanLine
from conf import getConfig, ANNEX_DIR
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
try:
wiki_dictionary = corpora.Dictionary.load_from_text(os.path.join(
getConfig('compass.gensim.training_data'), 'wiki_en_wordids.txt'))
wiki_corpus = corpora.MmCorpus(bz2.BZ2File(os.path.join(
getConfig('compass.gensim.training_data'), 'wiki_en_tfidf.mm.bz2')))
except Exception as e:
print "\n\n************** %s [ERROR] ******************\n" % task_tag
error_msg = "having trouble loading gensim dictionary and corpus from wiki dump: (error type %s)" % type(e)
print error_msg
print e
task.fail(message=error_msg)
return
from gensim import models
wiki_log_entropy_file = os.path.join(getConfig('compass.gensim.training_data'), 'wiki_en_log_entropy.model')
if not os.path.exists(wiki_log_entropy_file):
print "\n\n************** %s [WARN] ******************\n" % task_tag
print "no pre-prepared log entropy model. going to generate this here, now. might take a minute..."
logent_transformation = models.LogEntropyModel(wiki_corpus, id2word=wiki_dictionary)
logent_transformation.save(wiki_log_entropy_file)
else:
logent_transformation = models.LogEntropyModel.load(wiki_log_entropy_file)
tokenize_function = corpora.wikicorpus.tokenize
doc_corpus = [wiki_dictionary.doc2bow(tokenize_function(cleanLine(page).lower())) for page in texts]
doc_corpus = logent_transformation[doc_corpus]
wiki_tfidf_file = os.path.join(getConfig('compass.gensim.training_data'), 'wiki_en_tfidf.tfidf_model')
if not os.path.exists(wiki_tfidf_file):
print "\n\n************** %s [WARN] ******************\n" % task_tag
print "no pre-prepared tfidf model. going to generate this here, now. might take a minute..."
wiki_tfidf = models.TfidfModel(wiki_corpus)
wiki_tfidf.save(wiki_tfidf_file)
else:
wiki_tfidf = models.TfidfModel.load(wiki_tfidf_file)
doc_tfidf = wiki_tfidf[doc_corpus]
num_topics = 35
lsi = models.LsiModel(corpus=doc_tfidf, id2word=wiki_dictionary, num_topics=num_topics)
topics = []
t_lambda = lambda x : [float(x[0]), x[1]]
for t_group in [t.split("+") for t in [str(topic) for topic in lsi.print_topics(num_topics)]]:
topics.append([t_lambda(t.strip().replace('\"','').split("*")) for t in t_group])
lsi_topics = {
"topics" : topics,
"doc_comprehension" : []
}
doc_lsi = lsi[doc_tfidf]
for d in doc_lsi:
lsi_topics['doc_comprehension'].append(d)
topic_path = doc.addAsset(lsi_topics, "%s_topics.json" % doc.file_name, as_literal=False,
description="Gensim Topics dump (from LSI Model)", tags=[ASSET_TAGS["GM_TOPICS"]])
doc.addCompletedTask(task.task_path)
task.routeNext()
print "\n\n************** %s [END] ******************\n" % task_tag
task.finish()
def evaluateTextFile(task):
task_tag = "EVALUATING TEXT FILE"
print "\n\n************** %s [START] ******************\n" % task_tag
print "evaluating text file at %s" % task.doc_id
task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
media = UnveillanceDocument(_id=task.doc_id)
if media is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
if not media.queryFile(media.file_name):
print "NO DOCUMENT CONTENT"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
content = media.loadFile(media.file_name)
if content is None:
print "NO DOCUMENT CONTENT"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail()
return
from lib.Core.Utils.funcs import b64decode
un_b64 = b64decode(content)
# We have removed base 64-ing from the log files...
if un_b64 is None:
un_b64 = content
if un_b64 is not None:
from lib.Worker.Utils.funcs import getFileType
from vars import MIME_TYPES, MIME_TYPE_MAP
un_b64_mime_type = getFileType(un_b64, as_buffer=True)
if DEBUG:
print "MIME TYPE: %s" % un_b64_mime_type
if un_b64_mime_type not in [MIME_TYPES['pgp'], MIME_TYPES['wildcard']]:
err_msg = "MIME TYPE NOT USABLE"
print err_msg
print "\n\n************** %s [ERROR] ******************\n" % task_tag
task.fail(status=412, message=err_msg)
return
media.addAsset(un_b64, "%s.pgp" % media.file_name, description="un-b64'ed pgp asset")
media.addCompletedTask(task.task_path)
message_sentinel = "-----BEGIN PGP MESSAGE-----"
if un_b64[0:len(message_sentinel)] == message_sentinel:
task.put_next("PGP.decrypt.decrypt")
task.routeNext(inflate={
'pgp_file' : ".data/%s/%s.pgp" % (media._id, media.file_name)
})
task.finish()
print "\n\n************** %s [END] ******************\n" % task_tag
def getAssets(uv_task):
task_tag = "FETCHING DOCUMENTCLOUD ASSETS"
print "\n\n************** %s [START] ******************\n" % task_tag
print "getting DocumentCloud assets for %s" % uv_task.doc_id
uv_task.setStatus(412)
from lib.Worker.Models.cp_documentcloud_client import CompassDocumentCloudClient
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
document = UnveillanceDocument(_id=uv_task.doc_id)
if document is None:
print "\n\n************** %s [ERROR] ******************\n" % task_tag
print "Document is None"
return
if not hasattr(document, "dc_id"):
print "\n\n************** %s [ERROR] ******************\n" % task_tag
print "Document has not document cloud id!"
return
if not hasattr(uv_task, "auth_string"):
print "\n\n************** %s [ERROR] ******************\n" % task_tag
print "DocumentCloud upload needs an auth string"
return
dc_client = CompassDocumentCloudClient(auth_string=uv_task.auth_string)
dc_manifest = dc_client.download("documents/%s.json" % document.dc_id)
if dc_manifest is None:
print "\n\n************** %s [ERROR] ******************\n" % task_tag
print "No DocumentCloud manifest yet for %s." % document._id
return
document.addAsset(dc_manifest, "document_cloud_manifest.json", as_literal=False,
description="description of document on DocumentCloud",
tags=[ASSET_TAGS['DOC_CLOUD_MANIFEST'], ASSET_TAGS['DOC_CLOUD_DOC']])
dc_entities = dc_client.download("documents/%s/entities.json" % document.dc_id)
if dc_entities is None:
print "\n\n************** %s [WARN] ******************\n" % task_tag
print "No DocumentCloud entiteis yet for %s." % document._id
else:
entity_asset = document.addAsset(dc_entities, "document_cloud_entities.json",
as_literal=False, description="entites pulled from DocumentCloud",
tags=[ASSET_TAGS['DOC_CLOUD_ENTITIES'], ASSET_TAGS['DOC_CLOUD_DOC']])
from lib.Worker.Models.uv_text import UnveillanceText
if not hasattr(document, "text_id"):
text = UnveillanceText(inflate={
'file_name' : entity_asset,
'entities' : dc_entities['entities'],
'media_id' : document._id
})
document.text_id = text._id
document.save()
else:
text = UnveillanceText(_id=document.text_id)
text.entities = dc_entities['entities']
text.save()
document.addCompletedTask(uv_task.task_path)
uv_task.finish()
def j3mify(uv_task): task_tag = "J3MIFYING" print "\n\n************** %s [START] ******************\n" % task_tag print "j3mifying asset at %s" % uv_task.doc_id uv_task.setStatus(302) import os from lib.Worker.Models.uv_document import UnveillanceDocument from conf import DEBUG from vars import ASSET_TAGS media = UnveillanceDocument(_id=uv_task.doc_id) if media is None: print "DOC IS NONE" print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail() return j3m = media.loadAsset(uv_task.j3m_name) if j3m is None: error_message = "J3M IS NONE" print error_message print "\n\n************** %s [ERROR] ******************\n" % task_tag uv_task.fail(message=error_message) return import json print "JSSON HERE:" try: print type(j3m) j3m = json.loads(j3m) except Exception as e: print "\n\n************** J3MIFYING [WARN] ******************\n" print e print "json load once fail. trying again" print j3m if type(j3m) in [str, unicode]: try: j3m = json.loads(j3m) except Exception as e: print "\n\n************** J3MIFYING [WARN] ******************\n" print e print "json loads twice fail." print type(j3m) try: j3m_sig = j3m['signature'] except Exception as e: print "NO SIGNATURE TO EXTRACT" print "\n\n************** J3MIFYING [ERROR] ******************\n" uv_task.fail(status=412, message="No Signature in J3M.") return media.addAsset(j3m_sig, "j3m.sig", tags=[ASSET_TAGS['SIG']], description="The j3m's signature") media.addFile( media.addAsset(j3m['j3m'], "j3m.json", tags=[ASSET_TAGS['J3M']], description="The j3m itself.", as_literal=False), None, sync=True) media.addCompletedTask(uv_task.task_path) uv_task.j3m_name = "j3m.json" uv_task.save() uv_task.routeNext() uv_task.finish() print "\n\n************** %s [END] ******************\n" % task_tag
def generatePageMap(uv_task):
task_tag = "PAGE MAPPER"
print "\n\n************** %s [START] ******************\n" % task_tag
print "MAPPING PAGES FROM TEXT DOCUMENT at %s" % uv_task.doc_id
uv_task.setStatus(302)
from lib.Worker.Models.uv_document import UnveillanceDocument
from conf import DEBUG
from vars import ASSET_TAGS
doc = UnveillanceDocument(_id=uv_task.doc_id)
if doc is None:
print "DOC IS NONE"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
import os, json
try:
page_path = doc.getAssetsByTagName(ASSET_TAGS['TXT_JSON'])[0]['file_name']
pages = json.loads(doc.loadFile(os.path.join(doc.base_path, page_path)))
except Exception as e:
if DEBUG: print e
try:
bow_path = doc.getAssetsByTagName(ASSET_TAGS['BOW'])[0]['file_name']
bow = json.loads(doc.loadFile(os.path.join(doc.base_path, bow_path)))
except Exception as e:
if DEBUG: print e
if pages is None or bow is None:
print "NO PAGES OR BAG OF WORDS"
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
# with unique words in bag that are not stopwords
# for each page, word count of each
from numpy import intersect1d, setdiff1d
from conf import getConfig
if hasattr(uv_task, "stopwords"):
stopwords = uv_task.stopwords
else:
stopwords = os.path.join(getConfig('nlp_server.path'), "stopwords.json")
try:
with open(stopwords, 'rb') as S:
if hasattr(uv_task, "stopwords_lang"):
lang = uv_task.stopwords_lang
else:
lang = "english"
stopwords = json.loads(S.read())[lang]
except Exception as e:
print "NO STOPWORDS...\n%s" % e
print "\n\n************** %s [WARN] ******************\n" % task_tag
page_map = []
print "STOPWORDS: (len %d)\nTOP:\n%s\n" % (len(stopwords), stopwords[:10])
print "BAG OF WORDS: (len %d)\nTOP:\n%s\n" % (len(bow), bow[:10])
use_words = [w for w in setdiff1d(bow, stopwords).tolist() if len(w) > 1]
print "SIFTING BAG OF WORDS (old len: %d, new len: %d)" % (len(bow), len(use_words))
global_info = {}
for i, p in enumerate(pages):
if p is None: continue
page_bow = p.lower().split(" ")
words = intersect1d(use_words, page_bow).tolist()
if len(words) == 0: continue
map = []
frequency_max = 0
for word in words:
word_info = { 'word' : word, 'count' : page_bow.count(word) }
map.append(word_info)
if word_info['count'] > frequency_max: frequency_max = word_info['count']
if word not in global_info.keys():
global_info[word] = 0
global_info[word] += word_info['count']
page_map.append({ 'index' : i, 'map' : map, 'frequency_max' : frequency_max })
if len(page_map) > 0:
global_info['uv_page_map'] = page_map
asset_path = doc.addAsset(global_info, "page_map.json", as_literal=False,
description="word frequencies, page-by-page", tags=[ASSET_TAGS['PAGE_MAP']])
if asset_path is None or not doc.addFile(asset_path, None, sync=True):
print "COULD NOT SAVE ASSET."
print "\n\n************** %s [ERROR] ******************\n" % task_tag
uv_task.fail()
return
doc.addCompletedTask(uv_task.task_path)
uv_task.routeNext()
print "\n\n************** %s [END] ******************\n" % task_tag
uv_task.finish()
