def createHits(question, answers, params):
if SANDBOX:
mturk_url = 'mechanicalturk.sandbox.amazonaws.com'
preview_url = 'https://workersandbox.mturk.com/mturk/preview?groupId='
else:
mturk_url = 'mechanicalturk.amazonaws.com'
preview_url = 'https://mturk.com/mturk/preview?groupId='
#Create Hit Form Structure
overview = Overview()
overview.append_field('Title', 'We want to know the crowds opinion!')
overview.append(
FormattedContent(
'<a href="http://programthecrowd.com/">Visit us here</a>'))
questionContent = QuestionContent()
questionContent.append_field('Title', question)
answerChoices = SelectionAnswer(min=1,
max=1,
style='checkbox',
selections=answers,
type='text',
other=False)
q = Question(identifier='Help',
content=questionContent,
answer_spec=AnswerSpecification(answerChoices),
is_required=True)
questionForm = QuestionForm()
questionForm.append(overview)
questionForm.append(q)
hitIdList = []
global conn
# key = params['aws_access_key']
# secret = params['aws_secret_key']
conn = MTurkConnection(
aws_access_key_id='AKIAJBTEJI2RGTJH7OBA',
aws_secret_access_key='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',
host=mturk_url)
#For Loop to create and post hits
for i in range(0, NUMBER_OF_HITS):
create_hit_rs = conn.create_hit(questions=questionForm,
lifetime=LIFETIME,
max_assignments=NUMBER_OF_ASSIGNMENTS,
title=TITLE,
keywords=KEYWORDS,
reward=REWARD,
duration=DURATION,
approval_delay=APPROVAL_DELAY,
annotation=DESCRIPTION)
#print(preview_url + create_hit_rs[0].HITTypeId)
#print("HIT ID: " + create_hit_rs[0].HITId)
hitIdList.append(create_hit_rs[0].HITId)
return hitIdList
def createQuestionForm(overviewTitle, overviewDescription, numberOfTweets,
listOfTweets, listOfTweetIDs):
"""
Create an overview for an MTurk HIT
"""
#The Question Form should contain 1 overview and 3 odd questions
questionForm = QuestionForm()
#Define the Overview
overview = Overview()
Title = FormattedContent(overviewTitle)
overview.append(Title)
overviewDescription1 = FormattedContent(overviewDescription[0])
overviewDescription2 = FormattedContent(overviewDescription[1])
overviewDescription3 = FormattedContent(overviewDescription[2])
overview.append(overviewDescription1)
overview.append(overviewDescription2)
overview.append(overviewDescription3)
#Append the Overview to the Question Form
questionForm.append(overview)
#Create the Questions, and Add them
for i in xrange(numberOfTweets):
overview = Overview()
questionTitle = FormattedContent(
'<font face="Helvetica" size="2"><b> Tweet #' + str(i + 1) +
'</b></font>')
overview.append(questionTitle)
questionBody = FormattedContent('<font face="Helvetica" size="2">' +
listOfTweets[i] + '</font>')
overview.append(questionBody)
#answerTuple = tuple([('<a href="https://wikipedia.org/en/' + y.replace(" ","_") + '" target="_blank">' + y + "</a>") for y in list(listOfAnswers[i])])
#links = FormattedContent('<b>Links</b> | ' + answerTuple[0] + ' | ' + answerTuple[1])
#overview.append(links)
questionForm.append(overview)
question = createQuestion(listOfTweetIDs[i], i, listOfTweets[i],
["Positive", "Negative"])
questionForm.append(question)
return questionForm
def make_question(url, selections):
"""Build a formatted question to be posted on mechanical turk.
:param urls: Images to be displayed in mc question.
:param selections: Answer selections.
:param fta: True if the question includes a FreeTextAnswer.
"""
mc_overview = Overview()
mc_overview.append_field('Title', title)
question_form = QuestionForm()
question_form.append(mc_overview)
question_form.append(make_mc_question(url, selections))
return question_form
def __generate_qualification_test(self, question_data, num_correct, title):
'''
Returns a QuestionForm and AnswerKey for a qualification test from a list of sentence dictionaries
'''
# Get question and answer data
questions = map(
lambda (i, x): self.__generate_qualification_question(x, i),
enumerate(question_data))
answers = map(lambda (i, x): x["answer_key_" + str(i)],
enumerate(questions))
answer_key = self.__generate_answer_key(answers, num_correct,
len(question_data))
# Create form setup
qual_overview = Overview()
qual_overview.append_field("Title", title)
# Instructions
qual_overview.append(
FormattedContent("<h1>You must correctly code " +
str(num_correct) + " out of the " +
str(len(question_data)) +
" test sentences below.</h1>"))
qual_overview.append(
FormattedContent(
"<h2>Coding instructions are listed below. Please read through these carefully before continuing on to the coding task.</h2>"
))
inst_url = "https://s3.amazonaws.com/aws.drewconway.com/mt/experiments/cmp/html/instructions.html"
qual_overview.append(
FormattedContent(
'<iframe src="' + inst_url +
'" frameborder="0" width="1280" height="300" scrolling="auto">This text is necessary to ensure proper XML validation</iframe>'
))
# Create question form and append contents
qual_form = QuestionForm()
qual_form.append(qual_overview)
for q in questions:
i = q["question_num"]
qual_form.append(q["policy_area_" + str(i)])
qual_form.append(q["econ_scale_" + str(i)])
qual_form.append(q["soc_scale_" + str(i)])
return (qual_form, answer_key)
def create_HIT(mturk_conn,letter,imgur_links):
# Given a char and set of links
# create and push HIT
try:
canary = mturk_conn.get_account_balance()
except Exception as e1:
print "[Error Connecting]",e1
print "[Exiting]"
exit(1)
hit = None
#-HIT Properties
title = 'Select the Best Character'
description= ('Of the available options below, please select the best representation of the following chracter: '+letter+'\n Your vote will help determine which character gets selected to be used in a collaborative typeface.')
keywords = 'image, voting, opinions'
#-Question Overview
overview = Overview()
overview.append_field('Title', 'Choose the best looking letter')
#-Question
qc1 = QuestionContent()
qc1.append_field('Title','Select Letter')
# Generate Awnsers 1 per imgur_links[]
choices = boto_injector(imgur_links)
#-Awnser Choices
fta1 = SelectionAnswer(min=1, max=1,style='radiobutton',\
selections=choices,type='binary',other=False)
q1 = Question(identifier='design',content=qc1,\
answer_spec=AnswerSpecification(fta1),is_required=True)
#-Question Form
question_form = QuestionForm()
question_form.append(overview)
question_form.append(q1)
#Put the HIT up
try:
mturk_conn.create_hit(questions=question_form,max_assignments=5,title=title,description=description,keywords=keywords,duration = 60*HIT_TIME,reward=0.01)
print "Hit issued for item:",letter
except Exception as e1:
print "Could not issue hit",e1
def _gen_overview():
overview_title = 'Translate these sentences'
overview_content = """<p>Your task is to translate the Spanish sentences into English. Please make sure that your English translation:</p>
<ul>
<li>Is faithful to the original in both meaning and style</li>
<li>Is grammatical, fluent, and natural-sounding English</li>
<li>Does not add or delete information from the original text</li>
<li>Does not contain any spelling errors</li>
</ul>
<p>When creating your translation, please follow these guidelines:</p>
<ul>
<li><b>Do not use any machine translation systems (like transle.google.com)</b></li>
</ul>
"""
overview = Overview()
overview.append('Title', overview_title)
overview.append('FormattedContent', overview_content)
return overview
def make_free_text_question_form(questions):
if not questions:
raise ValueError('Questions cannot be empty!')
question_form = QuestionForm()
for q in questions:
qid = q['id']
for field in q['columns']:
if field == q['columns'][0]:
hint = SimpleField('Title', q['content'])
question_form.append(Overview([hint]))
field_id = str(qid) + free_sep + field
q_text = SimpleField('Text', field)
q_content = QuestionContent([q_text])
cons = Constraints([LengthConstraint(min_length=1, max_length=100)])
answer_spec = AnswerSpecification(FreeTextAnswer(constraints=cons))
question = Question(field_id, q_content, answer_spec, True)
question_form.append(question)
return question_form
def __generate_qualification_test(self, question_data, num_correct, title):
'''
Returns a QuestionForm and AnswerKey for a qualification test from a list of sentence dictionaries.
question_data : json object containing all the questions.
'''
# Get question and answer data
questions = map(
lambda (i, x): self.__generate_qualification_question(x, i),
enumerate(question_data))
answers = map(lambda (i, x): x["answer_key_" + str(i)],
enumerate(questions))
answer_key = self.__generate_answer_key(answers, num_correct,
len(question_data))
# Create form setup
qual_overview = Overview()
qual_overview.append_field("Title", title)
# Instructions
qual_overview.append(
FormattedContent(
"<h1>Please answer all the questions below.</h1>"))
qual_overview.append(
FormattedContent(
"<h2>For each question, please choose either the left or right image \
which you think is more beautiful in terms of its composition. Hints: Please make your decision based on\
several 'rules of thumb' in photography, such as rule of thirds, visual balance and golden ratio. \
You may also make your decision by judging which image contains less unimportant or distracting contents.</h2>"
))
# Create question form and append contents
qual_form = QuestionForm()
qual_form.append(qual_overview)
for q in questions:
i = q["question_num"]
qual_form.append(q["question_" + str(i)])
return (qual_form, answer_key)
def make_question_form_HIT(self,
audio_clip_urls,
hit_title,
question_title,
description,
keywords,
duration=DEFAULT_DURATION,
reward=DEFAULT_REWARD):
overview = Overview()
overview.append_field("Title", hit_title)
#overview.append(FormattedContent('<a target = "_blank" href="url">hyperlink</a>'))
question_form = QuestionForm()
question_form.append(overview)
for ac in audio_clip_urls:
audio_html = self.transcription_question.replace(
self.audio_url_tag, ac)
qc = QuestionContent()
qc.append_field("Title", question_title)
qc.append(FormattedContent(audio_html))
fta = FreeTextAnswer()
q = Question(identifier="transcription",
content=qc,
answer_spec=AnswerSpecification(fta))
question_form.append(q)
try:
response = self.conn.create_hit(questions=question_form,
max_assignments=1,
title=hit_title,
description=description,
keywords=keywords,
duration=duration,
reward=reward)
except MTurkRequestError as e:
if e.reason != "OK":
raise
return question_form, response
def validate():
print '\nValidating results on Mechanical Turk...\n'
ACCESS_ID = getpass.getpass("Enter your access ID: ")
SECRET_KEY = getpass.getpass("Enter your secret key: ")
reward, HOST, numWorkers, votes = getUserInput()
mtc = MTurkConnection(aws_access_key_id=ACCESS_ID,
aws_secret_access_key=SECRET_KEY,
host=HOST)
overview = Overview()
overview.append_field('Title', 'Select app features related to a set of words')
# Get features from file
questions = readFeatures()
# Create questions
mtc_questions = []
for q in questions:
mtc_questions.append(createQuestion(q[0], q[1]))
# Create survey with the questions inside
q_form = createForm(overview, mtc_questions)
# Post the survey
result = createHit(q_form, mtc, reward, numWorkers)
hitId = result[0].HITId
# Process the results when they come in
print '\nFetching results...\n'
data = fetchResults(mtc, questions, numWorkers, votes, hitId)
# Write the validated features to file
outputResults(data, numWorkers)
PIN_IMAGE_URL = 'http://media-cache-ak0.pinimg.com/236x/17/8f/99/178f993435fb2718ab6e22ba29d704e0.jpg'
PIN_IMAGE_TITLE = 'Arnhem Clothing \'Song Bird\' Kimono in Mayan Song. Via Soleilblue..'
mtc = MTurkConnection(aws_access_key_id=AWS_ACCESS_KEY_ID,
aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
host=HOST)
if debug: print mtc.get_account_balance()
title = 'Match these Pictures to Macy\'s Products'
description = 'Look at this photo and match it to Macy\'s products'
keywords = 'clothing, rating, opinions, easy, quick, macys'
#make overview
overview = Overview()
overview.append_field('Title',
'Find three Macys.com Product Web IDs That Match')
overview.append(
FormattedContent('<img src="' + PIN_IMAGE_URL + '" alt="Pintrest Image" />'
'<br />' + PIN_IMAGE_TITLE))
#make webid1
qc1 = QuestionContent()
qc1.append_field('Title', 'First WebID Code')
fta1 = FreeTextAnswer(num_lines=1)
q1 = Question(identifier="FirstWebCode",
content=qc1,
def submit_extract_keywords_hit(note):
"""Create a Mechanical Turk HIT that asks a worker to
choose keywords and definitions from the given note."""
MTURK_HOST = run_mturk('submit_extract_keywords_hit')
if not MTURK_HOST:
return
connection = MTurkConnection(settings.AWS_ACCESS_KEY_ID,
settings.AWS_SECRET_ACCESS_KEY,
host=MTURK_HOST)
if note.course.school:
title = KEYWORDS_HIT_TITLE_TEMPLATE.format(
course=note.course.name, school=note.course.school.name)
else:
title = KEYWORDS_HIT_TITLE_TEMPLATE.format(
course=note.course.name, school=note.course.department.school.name)
overview = Overview()
overview.append(
FormattedContent(
KEYWORDS_HIT_OVERVIEW_TEMPLATE.format(
domain=Site.objects.get_current(),
link=note.get_absolute_url())))
keyword_fta = FreeTextAnswer()
keyword_fta.num_lines = 1
definition_fta = FreeTextAnswer()
definition_fta.num_lines = 3
question_form = QuestionForm()
question_form.append(overview)
for i in range(
min(len(KEYWORDS_HIT_KEYWORD_FIELDS),
len(KEYWORDS_HIT_DEFINITION_FIELDS))):
keyword_content = QuestionContent()
keyword_content.append_field('Title',
KEYWORDS_HIT_KEYWORD_FIELDS[i][1])
keyword_question = Question(
identifier=KEYWORDS_HIT_KEYWORD_FIELDS[i][0],
content=keyword_content,
answer_spec=AnswerSpecification(keyword_fta),
is_required=True if i <= 10 else False)
question_form.append(keyword_question)
definition_content = QuestionContent()
definition_content.append_field('Title',
KEYWORDS_HIT_DEFINITION_FIELDS[i][1])
definition_question = Question(
identifier=KEYWORDS_HIT_DEFINITION_FIELDS[i][0],
content=definition_content,
answer_spec=AnswerSpecification(definition_fta),
is_required=False)
question_form.append(definition_question)
hit = connection.create_hit(questions=question_form,
max_assignments=1,
title=title,
description=KEYWORDS_HIT_DESCRIPTION,
keywords=KEYWORDS_HIT_KEYWORDS,
duration=KEYWORDS_HIT_DURATION,
reward=KEYWORDS_HIT_REWARD,
qualifications=KEYWORDS_HIT_QUALIFICATION,
annotation=str(note.id))[0]
KeywordExtractionHIT.objects.create(HITId=hit.HITId,
note=note,
processed=False)
def PublishTasks(hitNum, maxAssignments):
# sandbox in which to simulate: mechanicalturk.sandbox.amazonaws.com
# real environment: mechanicalturk.amazonaws.com
mtc = MTurkConnection(host='mechanicalturk.amazonaws.com')
# print mtc.APIVersion
# print mtc.get_account_balance()
# print mtc.get_reviewable_hits()
# print mtc.get_all_hits()
#--------------- BUILD OVERVIEW -------------------
# jbragg: Modified maximum reward description.
#title = '(Maximum reward possible: $70) Identify the relation between two entities in English sentences'
title = 'Identify the relation between two entities in English sentences'
#description = 'You will be given English sentences in which your task is to identify the relation between two designated entities. Your reward will depend on how many questions you have answered. The maximum reward you can earn is $70.'
description = 'You will be given English sentences in which your task is to identify the relation between two designated entities. Your reward will depend on how many questions you have answered. The maximum reward you can earn is approximately $5.'
keywords = 'English sentences, relation identification'
ratings = [('Very Bad', '-2'), ('Bad', '-1'), ('Not bad', '0'),
('Good', '1'), ('Very Good', '1')]
#--------------- BUILD OVERVIEW -------------------
overview = Overview()
overview_title = 'Exercise link (please copy the link and paste it in your browser if it cannot be opened directly.)'
link = '<a target="_blank"' ' href="http://128.208.3.167:3000/mturk">' ' http://128.208.3.167:3000/mturk</a>'
# jbragg: Commented out long-term bonus.
instructions = '<p>Instructions:</p><ul><li>You will be presented with sentences that have a person and a location highlighted.</li><li>Your task is to determine which of the 5 designated relations are expressed between the person and location.</li><li>You'll get paid $0.50 after each successful set of 20 questions<!-- -- plus a bonus of $2.00 after every 10 batches (equal to 200 questions)-->.</li><li>We know the correct answers to some of these sentence questions, and you can stay if you get these questions right.</li><li>You can start by going to the external link above now. After you finish all the questions, you will be provided with a confirm code, used for authentication and determining the appropriate amount of money as the payment.</li><li>In very rare cases where the website crashes, you could click backward and then forward on your browser to reload the question. It won\'t affect the payment because all the questions you have answered are recorded, on which the amount of payment is based. So please don\'t worry about that.</li></ul>'
overview_content = link + instructions
overview.append_field('Title', overview_title)
overview.append(FormattedContent(overview_content))
#--------------- BUILD QUESTION 1 -------------------
qc1 = QuestionContent()
qc1.append_field('Title', 'How looks the design ?')
fta1 = SelectionAnswer(min=1,
max=1,
style='dropdown',
selections=ratings,
type='text',
other=False)
q1 = Question(identifier='design',
content=qc1,
answer_spec=AnswerSpecification(fta1),
is_required=True)
#--------------- BUILD QUESTION 2 -------------------
qc2 = QuestionContent()
qc2.append_field(
'Title',
'Confirm code \n1. The code will be provided to you as you finish from the exercise link. \n2. The code will be verified before paying. \n3. By the end of every 20 questions (as a batch), You can choose to finish and get a confirm code, or continue.'
)
fta2 = FreeTextAnswer()
q2 = Question(identifier="Confirm_code",
content=qc2,
answer_spec=AnswerSpecification(fta2))
#--------------- BUILD THE QUESTION FORM -------------------
question_form = QuestionForm()
question_form.append(overview)
# question_form.append(q1)
question_form.append(q2)
#--------------- CREATE HITs -------------------
HIT_num = hitNum
for i in range(HIT_num):
# max_assignments: how many replicas this HIT has
mtc.create_hit(questions=question_form,
max_assignments=maxAssignments,
title=title,
description=description,
keywords=keywords,
duration=60 * 60 * 10,
reward=0.50)
def generate_hits(subset, begin_index, args, data_ids, images_metainfo):
from boto.mturk.connection import MTurkConnection
from boto.mturk.question import QuestionContent, Question, QuestionForm, Overview, AnswerSpecification, SelectionAnswer, FormattedContent, FreeTextAnswer
from boto.mturk.qualification import PercentAssignmentsApprovedRequirement, Qualifications
ACCESS_ID = amazon_config.ACCESS_ID
SECRET_KEY = amazon_config.SECRET_KEY
HOST = 'mechanicalturk.amazonaws.com'
mtc = MTurkConnection(aws_access_key_id=ACCESS_ID,
aws_secret_access_key=SECRET_KEY,
host=HOST)
title = 'Give your opinion of interestingness level about images'
description = (
'Watch images and give us your opinion of interestingness level about the images'
)
keywords = 'image, interestingness, interesting, rating, opinions'
ratings = [('Very boring', '-2'), ('Boring', '-1'), ('Neutral', '0'),
('Interesting', '1'),
('Very interesting, I would like to share it with my friends.',
'2')]
#--------------- BUILD OVERVIEW -------------------
overview = Overview()
overview.append_field(
'Title',
'Give your opinion about interestingness level on those images')
#--------------- BUILD QUESTIONs -------------------
questions = []
utils.write_file(subset, args.o + '.index_' + str(begin_index) + '.txt')
index = 0
for image_url in subset:
image_id = data_ids[index]
image = images_metainfo[image_id]
interestingness = 0
if ('repin_count' in image):
interestingness = int(image['repin_count']) + int(
image['like_count'])
#else:
# interestingness = int(image['interestingness'])
index = index + 1
qc = QuestionContent()
context = ''
if (interestingness > 0):
context = ' (shared by ' + str(interestingness) + ' people)'
qc.append_field('Title',
'How interesting the image' + context + ' to you?')
qc.append(
FormattedContent('<img src="' + image_url + '" alt="image" />'))
fta = SelectionAnswer(min=1,
max=1,
style='dropdown',
selections=ratings,
type='text',
other=False)
q = Question(identifier='interestingness',
content=qc,
answer_spec=AnswerSpecification(fta),
is_required=True)
questions.append(q)
#--------------- BUILD THE QUESTION FORM -------------------
question_form = QuestionForm()
question_form.append(overview)
for question in questions:
question_form.append(question)
# BUILD QUALIFICATION
qualifications = Qualifications()
req = PercentAssignmentsApprovedRequirement(comparator="GreaterThan",
integer_value="95")
qualifications.add(req)
#--------------- CREATE THE HIT -------------------
mtc.create_hit(questions=question_form,
qualifications=qualifications,
max_assignments=10,
title=title,
description=description,
keywords=keywords,
duration=60 * 30,
reward=0.2)
def check_notes_mailbox():
MTURK_HOST = run_mturk('get_extract_keywords_results')
if not MTURK_HOST:
return
try:
MAILBOX_USER = os.environ['NOTES_MAILBOX_USERNAME']
MAILBOX_PASSWORD = os.environ['NOTES_MAILBOX_PASSWORD']
FILEPICKER_API_KEY = os.environ['FILEPICKER_API_KEY']
except:
logger.warn(
'Could not find notes mailbox secrets, not running check_notes_mailbox'
)
return
connection = MTurkConnection(settings.AWS_ACCESS_KEY_ID,
settings.AWS_SECRET_ACCESS_KEY,
host=MTURK_HOST)
mailbox = poplib.POP3_SSL('pop.gmail.com', 995)
mailbox.user(MAILBOX_USER)
mailbox.pass_(MAILBOX_PASSWORD)
numMessages = len(mailbox.list()[1])
for i in range(numMessages):
# construct message object from raw message
raw_message_string = '\n'.join(mailbox.retr(i + 1)[1])
message = email.message_from_string(raw_message_string)
if not message.is_multipart():
logger.warn('Got an email with no attachments')
continue
attachments = []
message_body = ''
message_parts = message.get_payload()
for part in message_parts:
# Look for the message's plain text body
if part.get_content_type(
) == 'text/plain' and part['Content-Disposition'] is None:
message_body = part.get_payload()
# Look for attachments
elif part['Content-Disposition'] and 'attachment;' in part[
'Content-Disposition']:
attachment_mimetype = part.get_content_type()
attachment_filename = re.search(
CONTENT_DISPOSITION_REGEX,
part['Content-Disposition']).group('filename')
if part['Content-Transfer-Encoding'] == 'base64':
attachment_data = base64.decodestring(part.get_payload())
else:
attachment_data = part.get_payload()
# Upload attachment to filepicker
resp = requests.post('https://www.filepicker.io/api/store/S3?key={key}&policy={policy}&' \
'signature={signature}&mimetype={mimetype}&filename={filename}'
.format(key=FILEPICKER_API_KEY, policy=FP_POLICY_READ_WRITE,
signature=FP_SIGNATURE_READ_WRITE, mimetype=attachment_mimetype,
filename=attachment_filename),
data=attachment_data)
if resp.status_code == 200:
url = json.loads(resp.text)['url']
url = url + '?policy={policy}&signature={signature}'\
.format(policy=FP_POLICY_READ, signature=FP_SIGNATURE_READ)
attachments.append((url, attachment_filename))
else:
logger.warn('Could not upload an attachment to filepicker')
message_subject = message['Subject']
overview = Overview()
overview.append(
FormattedContent(
EMAIL_HIT_OVERVIEW_TEMPLATE.format(subject=message_subject,
body=message_body,
attachments='')))
single_line_answer = FreeTextAnswer()
single_line_answer.num_lines = 1
question_form = QuestionForm()
question_form.append(overview)
course_spam_content = QuestionContent()
course_spam_content.append_field(
'Title',
'Does the email contain course notes (check attachments below)?')
answer = SelectionAnswer(style='dropdown',
selections=[('No', 'no'), ('Yes', 'yes')])
course_spam = Question(identifier=COURSE_SPAM_QID,
content=course_spam_content,
answer_spec=AnswerSpecification(answer),
is_required=True)
question_form.append(course_spam)
course_name_content = QuestionContent()
course_name_content.append_field('Title', 'Course Name')
course_name = Question(
identifier=COURSE_NAME_QID,
content=course_name_content,
answer_spec=AnswerSpecification(single_line_answer),
is_required=True)
question_form.append(course_name)
instructor_names_content = QuestionContent()
instructor_names_content.append_field('Title', 'Instructor Name(s)')
instructor_names = Question(
identifier=INSTRUCTOR_NAMES_QID,
content=instructor_names_content,
answer_spec=AnswerSpecification(single_line_answer),
is_required=False)
question_form.append(instructor_names)
school_name_content = QuestionContent()
school_name_content.append_field('Title', 'School Name')
school_name = Question(
identifier=SCHOOL_NAME_QID,
content=school_name_content,
answer_spec=AnswerSpecification(single_line_answer),
is_required=True)
question_form.append(school_name)
department_name_content = QuestionContent()
department_name_content.append_field('Title', 'Department Name')
department_name = Question(
identifier=DEPARTMENT_NAME_QID,
content=department_name_content,
answer_spec=AnswerSpecification(single_line_answer),
is_required=False)
question_form.append(department_name)
for i in range(len(attachments)):
overview = Overview()
overview.append(
FormattedContent(
EMAIL_HIT_ATTACHMENT_OVERVIEW_TEMPLATE.format(
link=attachments[i][0], name=attachments[i][1])))
question_form.append(overview)
note_title_content = QuestionContent()
note_title_content.append_field('Title', 'Note Title')
note_title = Question(
identifier=NOTE_TITLE_QID_TEMPLATE + str(i),
content=note_title_content,
answer_spec=AnswerSpecification(single_line_answer),
is_required=True)
question_form.append(note_title)
note_category_content = QuestionContent()
note_category_content.append_field('Title', 'Note Category')
answer = SelectionAnswer(style='dropdown',
selections=NOTE_CATEGORIES_FOR_MTURK)
note_category = Question(identifier=NOTE_CATEGORY_QID_TEMPLATE +
str(i),
content=note_category_content,
answer_spec=AnswerSpecification(answer),
is_required=True)
question_form.append(note_category)
hit = connection.create_hit(questions=question_form,
max_assignments=1,
title=EMAIL_HIT_TITLE,
description=EMAIL_HIT_DESCRIPTION,
keywords=EMAIL_HIT_KEYWORDS,
duration=EMAIL_HIT_DURATION,
reward=EMAIL_HIT_REWARD,
qualifications=EMAIL_HIT_QUALIFICATION)[0]
def make_html_elicitation_HIT(self,
prompt_list,
hit_title,
prompt_title,
keywords,
hit_description,
duration=DEFAULT_DURATION,
reward_per_clip=DEFAULT_REWARD,
max_assignments=DEFAULT_MAX_ASSIGNMENTS):
overview = Overview()
overview.append_field(
"Title", "Record yourself speaking the words in the prompt.")
descriptions = [
"The following prompts are in English.",
"Click the prompt to record your voice (Redirects to recording Page).",
"Follow the directions on that page.",
"Copy and paste the URL in box below the prompt on this page."
]
keywords = "audio, recording, elicitation, English"
html_head = self.elicitation_head.replace(self.html_tags["title"],
hit_title)
for description in descriptions:
html_head = html_head.replace(
self.html_tags["description"], "<li>" + description +
"</li>\n" + self.html_tags["description"])
questions_html = []
prompt_ids = []
for prompt_words, prompt_id in prompt_list:
#For each prompt, generate the question html given the template
prompt_id = str(prompt_id)
prompt = " ".join(prompt_words)
underscored_prompt = "_".join(prompt_words)
question = self.elicitation_question.replace(
self.html_tags["prompt"], prompt)
question = question.replace(self.html_tags["underscored_prompt"],
underscored_prompt)
question = question.replace(self.html_tags["prompt_id"],
str(prompt_id))
questions_html.append(question)
prompt_ids.append(prompt_id)
for prompt_id in prompt_ids:
#Disable the inputs for the prompts, which are just text fields for the
#audio recording URLs
script = self.disable_input_script.replace("${input_id}",
prompt_id)
html_head = html_head.replace(self.html_tags["disable_script"],script+\
"\n"+self.html_tags["disable_script"])
if (self.html_tags["prompt_id"]) in html_head:
html_head = html_head.replace(self.html_tags["prompt_id"],"'"+prompt_id+"'"+\
","+self.html_tags["prompt_id"])
#Get rid of html tags
html_head = html_head.replace(self.html_tags["disable_script"], "")
html_head = html_head.replace("," + self.html_tags["prompt_id"], "")
html_head = html_head.replace(self.html_tags["description"], "")
html = html_head
for question in questions_html:
html += question
html += self.transcription_tail
html_question = HTMLQuestion(html, 800)
open("/home/taylor/csaesr/tmp/hithtml.html", "w").write(html)
#reward calculation
quals = qualification.Qualifications()
quals.add(qualification.LocaleRequirement("EqualTo", "US"))
reward = reward_per_clip * len(prompt_list)
try:
return self.conn.create_hit(title=hit_title,
question=html_question,
max_assignments=max_assignments,
description=hit_description,
qualifications=quals,
keywords=keywords,
duration=duration,
reward=reward)
except MTurkRequestError as e:
if e.reason != "OK":
raise
else:
print(e)
return False
return False
def new_rate_hit(PIN_IMAGE_URL, PIN_IMAGE_TITLE, MACYS_IMAGE_URL,
MACYS_IMAGE_TITLE):
mtc = MTurkConnection(aws_access_key_id=AWS_ACCESS_KEY_ID,
aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
host=HOST)
if debug: print mtc.get_account_balance()
title = 'Match these Pictures to Macy\'s Products'
description = 'Look at this photo and match it to Macy\'s products'
keywords = 'clothing, rating, opinions, easy, quick, macys'
ratings = [('Very Bad', '1'), ('Bad', '2'), ('OK', '3'), ('Good', '4'),
('Very Good', '5')]
#make overview
overview = Overview()
overview.append_field('Title', 'Rank how these two images match.')
overview.append(
FormattedContent('<table border="1">><tr><td width="50%"><img src="' +
PIN_IMAGE_URL + '" alt="Pintrest Image" /></td>'
'<td width="50%"><img src="' + MACYS_IMAGE_URL +
'" alt="Macys Image" /></td></tr><tr>'
'<td width="50%">' + PIN_IMAGE_TITLE +
'</td><td width="50%">' + MACYS_IMAGE_TITLE +
'</td></tr></table>'))
#make q1
qc1 = QuestionContent()
qc1.append_field('Title', 'Rank the match between these two')
fta1 = SelectionAnswer(min=1,
max=1,
style='dropdown',
selections=ratings,
type='text',
other=False)
q1 = Question(identifier='rating',
content=qc1,
answer_spec=AnswerSpecification(fta1),
is_required=True)
#make q2
qc2 = QuestionContent()
qc2.append_field('Title', 'Comments about the HIT (Optional)')
fta2 = FreeTextAnswer()
q2 = Question(identifier="comments",
content=qc2,
answer_spec=AnswerSpecification(fta2))
#make question form
question_form = QuestionForm()
question_form.append(overview)
question_form.append(q1)
question_form.append(q2)
#--------------- CREATE THE HIT -------------------
mtc.create_hit(questions=question_form,
max_assignments=1,
title=title,
description=description,
keywords=keywords,
duration=60 * 5,
reward=0.05)
def new_sugg_hit(PIN_IMAGE_URL, PIN_IMAGE_TITLE):
mtc = MTurkConnection(aws_access_key_id=AWS_ACCESS_KEY_ID,
aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
host=HOST)
if debug: print mtc.get_account_balance()
title = 'Match these Pictures to Macy\'s Products'
description = 'Look at this photo and match it to Macy\'s products'
keywords = 'clothing, rating, opinions, easy, quick, macys'
#make overview
overview = Overview()
overview.append_field('Title',
'Find three Macys.com Product Web IDs That Match')
overview.append(
FormattedContent('<img src="' + PIN_IMAGE_URL +
'" alt="Pintrest Image" />'
'<br />' + PIN_IMAGE_TITLE))
#make webid1
qc1 = QuestionContent()
qc1.append_field('Title', 'First WebID Code')
fta1 = FreeTextAnswer(num_lines=1)
q1 = Question(identifier="FirstWebCode",
content=qc1,
answer_spec=AnswerSpecification(fta1))
#make webid2
qc2 = QuestionContent()
qc2.append_field('Title', 'Second WebID Code')
fta2 = FreeTextAnswer(num_lines=1)
q2 = Question(identifier="SecondWebCode",
content=qc2,
answer_spec=AnswerSpecification(fta2))
#make webid1
qc3 = QuestionContent()
qc3.append_field('Title', 'Third WebID Code')
fta3 = FreeTextAnswer(num_lines=1)
q3 = Question(identifier="ThirdWebCode",
content=qc3,
answer_spec=AnswerSpecification(fta3))
#make question form
question_form = QuestionForm()
question_form.append(overview)
question_form.append(q1)
question_form.append(q2)
question_form.append(q3)
#--------------- CREATE THE HIT -------------------
mtc.create_hit(questions=question_form,
max_assignments=1,
title=title,
description=description,
keywords=keywords,
duration=60 * 5,
reward=0.05)
def demo_create_favorite_color_hit():
"""A HIT to determine the Worker's favorite color"""
TITLE = 'Tell me your favorite color'
DESCRIPTION = ('This is a HIT that is created by a computer program '
'to demonstrate how Mechanical Turk works. This should '
'be a free HIT for the worker.')
KEYWORDS = 'data collection, favorite, color'
DURATION = 15 * 60 # 15 minutes (Time to work on HIT)
MAX_ASSIGNMENTS = 1 # Number of assignments per HIT
REWARD_PER_ASSIGNMENT = 0.00 # $0.00 USD (1 cent)
#--------------- BUILD HIT container -------------------
overview = Overview()
overview.append_field('Title', TITLE)
overview.append(
FormattedContent(
"<p>This is an experiment to learn Mechanical Turk</p>"))
#--------------- BUILD QUESTION 1 -------------------
question_content = QuestionContent()
question_content.append(
FormattedContent(
"<b>What is your favorite color?</b> There isn't a financial "
"reward for answering, but you will get an easy approval for your "
"statistics."))
free_text_answer = FreeTextAnswer(num_lines=1)
q1 = Question(identifier='favorite_color',
content=question_content,
answer_spec=AnswerSpecification(free_text_answer),
is_required=True)
#--------------- BUILD QUESTION 3 -------------------
question_content = QuestionContent()
question_content.append(
FormattedContent("""<p>Give me a fun comment:</p>"""))
q2 = Question(identifier="comments",
content=question_content,
answer_spec=AnswerSpecification(FreeTextAnswer()))
#--------------- BUILD THE QUESTION FORM -------------------
question_form = QuestionForm()
question_form.append(overview)
question_form.append(q1)
question_form.append(q2)
#--------------- CREATE THE HIT -------------------
mtc = get_connection()
hit = mtc.create_hit(questions=question_form,
max_assignments=MAX_ASSIGNMENTS,
title=TITLE,
description=DESCRIPTION,
keywords=KEYWORDS,
duration=DURATION,
reward=REWARD_PER_ASSIGNMENT)
#---------- SHOW A LINK TO THE HIT GROUP -----------
base = get_worker_url()
print "\nVisit this website to see the HIT that was created:"
print "%s/mturk/preview?groupId=%s" % (base, hit[0].HITTypeId)
return hit[0]
def create_crop_hit(mturk,
URLs,
num_assignment,
qualification=Qualifications()):
# Constant data for HIT generation
hit_title = "Photo Quality Assessment"
hit_description = "This task involves viewing pairs of pictures and judging which picture among the image pair is more beautiful."
lifetime = 259200
keywords = ["photo", "quality", "ranking"]
duration = 30 * 60
reward = 0.05
#approval_delay = 86400
# Question form for the HIT
question_form = QuestionForm()
overview = Overview()
overview.append_field('Title', 'Photo Quality Assessment')
overview.append(
FormattedContent(
'For each question, please choose either the left or right image which you think is more beautiful in terms of its <u>composition</u>.'
))
overview.append(
FormattedContent(
'<b>Hints: Please make your decision based on several "rules of thumb" in photography, such as rule of thirds, visual balance and golden ratio.</b>'
))
#overview.append(FormattedContent('<b>You may also make your decision by judging which image contains less unimportant or distracting contents</b>.'))
overview.append(
FormattedContent(
'For those hard cases, please just select your preferred image based on your sense of aesthetics.'
))
question_form.append(overview)
ratings = [('Left', '0'), ('Right', '1')]
for i in xrange(len(URLs)):
qc = QuestionContent()
qc.append_field('Title', 'Question')
qc.append_field(
'Text',
'Please indicate which one of the following images is more beautiful.'
)
qc.append(
FormattedContent('<img src="' + URLs[i] +
'" alt="Image not shown correctly!"></img>'))
#URLs[i]
fta = SelectionAnswer(min=1,
max=1,
style='radiobutton',
selections=ratings,
type='text',
other=False)
q = Question(identifier='photo_pair_' + str(i),
content=qc,
answer_spec=AnswerSpecification(fta),
is_required=True)
question_form.append(q)
hit_res = mturk.create_hit(
title=hit_title,
description=hit_description,
reward=Price(amount=reward),
duration=duration,
keywords=keywords,
#approval_delay=approval_delay,
question=question_form,
#lifetime=lifetime,
max_assignments=num_assignment,
qualifications=qualification)
# return HIT ID
return hit_res[0].HITId
def make_html_transcription_HIT(self,
audio_clip_urls,
hit_title,
question_title,
description,
keywords,
duration=DEFAULT_DURATION,
reward_per_clip=DEFAULT_REWARD,
max_assignments=DEFAULT_MAX_ASSIGNMENTS):
overview = Overview()
overview.append_field(
"Title", "Type the words in the following audio clip in order.")
descriptions = [
"The following audio clips are in English.",
"Transcribe the audio clip by typing the words that the person \
says in order.",
"Do not use abbreviations: 'street' and NOT 'st.'",
"Write numbers long-form, as in: 'twenty fifth' NOT '25th'.",
"Write letters (see example).", "Punctuation does not matter.",
"Hotkeys: press Tab to play the next clip."
]
keywords = "audio, transcription, English"
html_head = self.transcription_head.replace(self.html_tags["title"],
hit_title)
for description in descriptions:
html_head = html_head.replace(
self.html_tags["description"], "<li>" + description +
"</li>\n" + self.html_tags["description"])
count = 0
questions = []
inputs = []
for acurl, acid in audio_clip_urls:
input_id = str(acid)
question = self.transcription_question.replace(
self.html_tags["audio_url"], acurl)
question = question.replace(self.html_tags["audioclip_id"],
str(acid))
question = question.replace("${count}", input_id)
count += 1
questions.append(question)
inputs.append(input_id)
for input_id in inputs:
script = self.disable_input_script.replace("${input_id}", input_id)
html_head = html_head.replace(self.html_tags["disable_script"],script+\
"\n"+self.html_tags["disable_script"])
if (self.html_tags["audio_id"]) in html_head:
html_head = html_head.replace(self.html_tags["audio_id"],"'"+\
input_id+"'"+","+self.html_tags["audio_id"])
html_head = html_head.replace(self.html_tags["disable_script"], "")
html_head = html_head.replace("," + self.html_tags["audio_id"], "")
html_head = html_head.replace(self.html_tags["description"], "")
html = html_head
for question in questions:
html += question
count += 1
html += self.transcription_tail
html_question = HTMLQuestion(html, 800)
#reward calculation
reward = reward_per_clip * len(audio_clip_urls)
try:
return self.conn.create_hit(title=hit_title,
question=html_question,
max_assignments=max_assignments,
description=description,
keywords=keywords,
duration=duration,
reward=reward)
except MTurkRequestError as e:
if e.reason != "OK":
raise
else:
print(e)
return False
return False
def generate_hits(mtc_type, subset, begin_index, args):
from boto.mturk.connection import MTurkConnection
from boto.mturk.question import QuestionContent, Question, QuestionForm, Overview, AnswerSpecification, SelectionAnswer, FormattedContent, FreeTextAnswer
from boto.mturk.qualification import PercentAssignmentsApprovedRequirement, Qualifications, Requirement
mtc = mtk_utils.get_mtc(mtc_type)
title = 'Give your opinion of aesthetics level about images'
description = ('View images and give us your opinion of aesthetics level about the images')
keywords = 'image, aesthetic, aesthetics, rating, opinions'
ratings =[('Very ugly','-2'),
('Ugly','-1'),
('Neutral','0'),
('Beautiful','1'),
('Very beautiful, I would like to take such beautiful photo too.','2')]
#--------------- BUILD OVERVIEW -------------------
overview = Overview()
overview.append_field('Title', 'Give your opinion about aesthetics level on those images')
#--------------- BUILD QUESTIONs -------------------
questions = []
if (args.m != 'qua'):
utils.write_file(subset, args.o + '.index_' + str(begin_index) + '.txt')
if (args.m == 'qua_init' and begin_index > 0):
return
for image_url in subset:
qc = QuestionContent()
qc.append_field('Title','How beautiful the image to you?')
qc.append(FormattedContent('<img src="' + image_url + '" alt="image" />'))
fta = SelectionAnswer(min=1, max=1,style='dropdown',
selections=ratings,
type='text',
other=False)
q = Question(identifier='aesthetics',
content=qc,
answer_spec=AnswerSpecification(fta),
is_required=True)
questions.append(q)
#--------------- BUILD THE QUESTION FORM -------------------
question_form = QuestionForm()
question_form.append(overview)
for question in questions:
question_form.append(question)
# BUILD QUALIFICATION
qualifications = Qualifications()
req = PercentAssignmentsApprovedRequirement(comparator = "GreaterThan", integer_value = "95")
qualifications.add(req)
if (args.m == 'qua'):
if (args.q != None):
qua_req = Requirement(qualification_type_id = args.q, comparator = 'EqualTo', integer_value = '1')
qualifications.add(qua_req)
else:
print("Please give qualification type id in 'qua' mode.")
sys.exit(0)
#--------------- CREATE THE HIT -------------------
hit = mtc.create_hit(questions = question_form,
qualifications = qualifications,
max_assignments = 10 * 2,
title = title,
description = description,
keywords = keywords,
duration = 60 * 30 * 2,
reward = 0.2 * 2)
if (args.m == 'qua_init'):
print("Create qualification type for HIT id: " + hit[0].HITId)
quatype = mtc.create_qualification_type(name = hit[0].HITId, description = "Temporary qualification for HIT " + hit[0].HITId, status = 'Active')
print("Qualification type id: " + quatype[0].QualificationTypeId)
def make_hit(image_url):
title = 'Label image with its location'
description = 'Answer questions about an image to label its location.'
keywords = 'image categorization, locations, scene recognition'
in_out = [('indoors', '0'), ('outdoors', '1')]
nat_manmade = [('man-made', '0'), ('natural', '1')]
functions = [('transportation/urban', '0'), ('restaurant', '1'),
('recreation', '2'), ('domestic', '3'),
('work/education', '4'), ('other/unclear', '5')]
landscapes = [('body of water/beach', '0'), ('field', '1'),
('mountain', '2'), ('forest/jungle', '3'),
('other/unclear', '4')]
#--------------- BUILD OVERVIEW -------------------
overview = Overview()
overview.append_field('Title', title)
with open(INSTRUCTIONS_HTML) as html:
instructions = html.read()
overview.append(FormattedContent(instructions))
image = Binary('image', None, image_url, 'image')
overview.append(image)
#--------------- BUILD QUESTION 1 -------------------
qc1 = QuestionContent()
qc1.append_field(
'Text', 'Is the location shown in the image indoors or outdoors?')
fta1 = SelectionAnswer(min=1,
max=1,
style='checkbox',
selections=in_out,
type='text',
other=False)
q1 = Question(identifier='Question 1',
content=qc1,
answer_spec=AnswerSpecification(fta1),
is_required=True)
#--------------- BUILD QUESTION 2 -------------------
qc2 = QuestionContent()
qc2.append_field(
'Text', 'Is the location shown in the image man-made or ' +
'natural? Examples of man-made locations include ' +
'buildings and parks while examples of natural ' +
'locations include mountains and rivers.')
fta2 = SelectionAnswer(min=1,
max=1,
style='checkbox',
selections=nat_manmade,
type='text',
other=False)
q2 = Question(identifier='Question 2',
content=qc2,
answer_spec=AnswerSpecification(fta2),
is_required=True)
#--------------- BUILD QUESTION 3 -------------------
qc3 = QuestionContent()
qc3.append_field(
'Text', 'If the location in the image is man-made, what is the ' +
'general function or type of the location? If the ' +
'location is natural (not man-made), don\'t select ' +
'anything here.')
fta3 = SelectionAnswer(min=0,
max=1,
style='checkbox',
selections=functions,
type='text',
other=False)
q3 = Question(identifier='Question 3',
content=qc3,
answer_spec=AnswerSpecification(fta3),
is_required=False)
#--------------- BUILD QUESTION 4 -------------------
qc4 = QuestionContent()
qc4.append_field(
'Text', 'If the location in the picture is natural, what ' +
'kind of natural location is it? If the location ' +
'man-made (not natural), don\'t select anything here.')
fta4 = SelectionAnswer(min=0,
max=1,
style='checkbox',
selections=landscapes,
type='text',
other=False)
q4 = Question(identifier='Question 4',
content=qc4,
answer_spec=AnswerSpecification(fta4),
is_required=False)
#--------------- BUILD THE QUESTION FORM -------------------
question_form = QuestionForm()
question_form.append(overview)
question_form.append(q1)
question_form.append(q2)
question_form.append(q3)
question_form.append(q4)
#-------------- QUALIFICATIONS -------------------
percent = PercentAssignmentsApprovedRequirement('GreaterThanOrEqualTo', 95)
number = NumberHitsApprovedRequirement('GreaterThanOrEqualTo', 200)
quals = Qualifications()
quals.add(percent)
quals.add(number)
#--------------- CREATE THE HIT -------------------
mtc.create_hit(questions=question_form,
max_assignments=1,
title=title,
description=description,
keywords=keywords,
qualifications=quals,
annotation=image_url,
duration=60 * 10,
reward=0.03)
def make_question_form_elicitation_HIT(
self,
prompt_list,
hit_title,
prompt_title,
keywords,
duration=DEFAULT_DURATION,
reward_per_clip=DEFAULT_REWARD,
max_assignments=DEFAULT_MAX_ASSIGNMENTS):
overview = Overview()
overview.append_field("Title", hit_title)
#overview.append(FormattedContent('<a target = "_blank" href="url">hyperlink</a>'))
question_form = QuestionForm()
descriptions = [
"The following prompts are in English.",
"Approve the flash permissions to record audio.",
"Click the red circle to record yourself.",
"Read the words after 'prompt:'", "Click 'Click to Stop'",
"Play the clip back to verify sound quality.",
"After you are happy with your recording, click 'Click here to save >>'",
"Copy & paste the URL under 'Sharing options' into the text field for the prompt.",
"You will NEVER be asked to divulge any personal or identifying information."
]
keywords = "audio, recording, elicitation, English"
# for i, description in enumerate(descriptions):
# overview.append_field("%dDescription"%i, description)
# flash_xml = FlashXml(self.flash_xml.replace(self.html_tags["flash_url"],self.vocaroo_url))
# overview.append(flash_xml)
question_form.append(overview)
qc = QuestionContent()
# qc.append(FormattedContent(flash_xml))
qc.append_field("Title",
"Please select the type of microphone you are using.")
# qc.append(Flash(self.vocaroo_url,525,450))
#
#answer = FreeTextAnswer()
answer = SelectionAnswer(max=1,
style="radiobutton",
selections=self.mic_selections)
q = Question(identifier="MIC",
content=qc,
answer_spec=AnswerSpecification(answer))
question_form.append(q)
qual = qualification.LocaleRequirement("in", "USA")
reward = reward_per_clip * len(prompt_list)
xml = question_form.get_as_xml()
try:
response = self.conn.create_hit(questions=question_form,
max_assignments=1,
title=hit_title,
qualification=qual,
description=descriptions[0],
keywords=keywords,
duration=duration,
reward=reward)
except MTurkRequestError as e:
if e.reason != "OK":
raise
return True
