CrowdFlower API - Python client
Python Makefile
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
crowdflower
examples
.gitignore
LICENSE
MANIFEST.in
Makefile
README.md
README.rst
setup.py

README.md

crowdflower

Client library for interacting with the CrowdFlower API with Python.

Installation

Install from PyPI with setuptools:

easy_install -U crowdflower

Or with pip:

pip install -U crowdflower

Or install the latest (potentially unreleased and unstable) code from GitHub, using pip:

git+https://github.com/twosigma/ngrid

Or build the source yourself, with setuptools:

git clone https://github.com/peoplepattern/crowdflower.git
cd crowdflower
python setup.py develop

Basic usage

Import like:

import crowdflower

CrowdFlower API keys are 20 characters long; the one below is just random characters. (You can find your API key at make.crowdflower.com/account/user.)

conn = crowdflower.Connection(api_key='LbcxvIlE3x1M8F6TT5hN')

The library will default to an environment variable called CROWDFLOWER_API_KEY if none is specified here:

conn = crowdflower.Connection()

If you want to cache job responses, like judgments, properties, and tags, you can initialize the connection with a cache. cache='filesystem' is the only option currently supported, and serializes JSON files to /tmp/crowdflower/*.

conn = crowdflower.Connection(cache='filesystem')

Inspecting existing jobs

Loop through all your jobs and print the titles:

for job in conn.jobs():
    print job.properties['title']

Creating a new job

Create a new job with some new units:

data = [
    {'id': '1', 'name': 'Chris Narenz', 'gender_gold': 'male'},
    {'id': '2', 'name': 'George Henckels'},
    {'id': '3', 'name': 'Maisy Ester'},
]
job = conn.upload(data)
update_result = job.update({
    'title': 'Gender labels',
    'included_countries': ['US', 'GB'],  # Limit to the USA and United Kingdom
        # Please note, if you are located in another country and you would like
        # to experiment with the sandbox (internal workers) then you also need
        # to add your own country. Otherwise your submissions as internal worker
        # will be rejected with Error 301 (low quality).
    'payment_cents': 5,
    'judgments_per_unit': 2,
    'instructions': 'some <i>instructions</i> html',
    'cml': 'some layout cml, e.g., '
        '<cml:text label="Sample text field:" validates="required" />',
    'options': {
        'front_load': 1, # quiz mode = 1; turn off with 0
    }
})

if 'errors' in update_result:
    print(update_result['errors'])
    exit()

job.gold_add('gender', 'gender_gold')

Launch job for on-demand workers (the default):

job.launch(2)

Launch job for internal workers (sandbox):

job.launch(2, channels=['cf_internal'])

Check the status of the job:

print job.ping()

Clean up; delete all the jobs that were created by the above example:

for job in conn.jobs():
    if job.properties['title'] == 'Gender labels':
        print 'Deleting Job#%s' % job.id
        print job.delete()

View annotations collected so far:

for row in job.download():
    print row

Example

See the README.md in the examples/ directory for a full spam classification example using the freely available SMS Spam Collection.

Debugging / Logging

To turn on verbose logging use the following in your script:

import logging
logging.basicConfig(level=logging.DEBUG)

Motivation

The official Ruby client is hard to use, which is surprising, since the CrowdFlower API is so simple.

Which is not to say the CrowdFlower API is all ponies and rainbows, but all the documentation is there on one page, and it does what it says, for the most part. (Though there's more that you can do, beyond what's documented.)

Thus, a thin Python client for the CrowdFlower API.

References

The CrowdFlower blog is the definitive (but incomplete) source for API documentation:

The source code for the official ruby-crowdflower project is also helpful in some cases.

This package uses kennethreitz's Requests to communicate with the CrowdFlower API over HTTP. Requests is Apache2 licensed.

Support

Found a bug? Want a new feature? File an issue!

Contributing

We love open source and working with the larger community to make our codebase even better! If you have any contributions, please fork this repository, commit your changes to a new branch, and then submit a pull request back to this repository (peoplepattern/crowdflower). To expedite merging your pull request, please follow the stylistic conventions already present in the repository. These include:

  • Adhere to PEP8
    • We're not super strict on every single PEP8 convention, but we have a few hard requirements:
      • Four-space indentation
      • No tabs
      • No semicolons
      • No wildcard imports
  • No trailing whitespace
  • Use docstrings liberally

The Apache License 2.0 contains a clause covering the Contributor License Agreement.

Authors

License

Copyright 2014 People Pattern Corporation

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.