Python 2 vs 3

DLATK is available for python 2.7 and 3.5, with the 3.5 version being the official release. The 2.7 version is fully functional (as of v0.6.1) but will not be maintained and also does not contain some of the newer features available in v1.0. Please make sure you pip is bound to Python 3 for the commands below.

To install the Python 2.7 version run:

pip install "dlatk < 1.0"

# install DLATK locally (no sudo permissions needed)
pip install --user "dlatk < 1.0"

Full Install

Setup (Linux)

WARNING: This will install MySQL on your computer.

Install the required Ubuntu libraries. The requirements.sys can be found on the DLATK GitHub page. The r-base package might be difficult to install and can be removed from requirements.sys if needed though this will limit some functionality.

xargs apt-get install < requirements.sys

DLATK has been tested on Ubuntu 14.04.

Setup (OSX with brew)

WARNING: This will install MySQL on your computer.

Install dependencies with brew.

brew install python mysql

DLATK has been tested on OSX 10.11.

Install (pip)

Install the Python 3 version via pip:

pip install dlatk

To install the Python 2.7 version use:

pip install "dlatk < 1.0"

Install (Anaconda)

Run the following in a Python 3.5 conda env:

conda install -c wwbp dlatk

Install (GitHub)

Run the following:

git clone
cd dlatk
python install

Install Sample Datasets

DLATK comes packaged with two sample databases: dla_tutorial and permaLexicon. See Packaged Datasets for more information on the databases. To install them use the following:

mysql -u username -p  < /path/to/dlatk/data/dla_tutorial.sql
mysql -u username -p  < /path/to/dlatk/data/permaLexicon.sql

The path to DLATK can be found using the following:

python -c "import dlatk; print(dlatk.__file__)"

WARNING: if these databases already exist the above commands will add tables to the db.

Install Other Dependencies

Load NLTK corpus

Load NLTK data from the command line:

python -c "import nltk;'wordnet')"

Install Stanford Parser

  1. Download the zip file from
  2. Extract into ../dlatk/Tools/StanfordParser/.
  3. Move ../dlatk/Tools/StanfordParser/ into the folder you extracted: ../dlatk/Tools/StanfordParser/stanford-parser-full*/.

Install Tweet NLP v0.3 (ark-tweet-nlp-0.3)

  1. Download the tgz file (for version 0.3) from
  2. Extract this file into ../dlatk/Tools/TwitterTagger/.

Python Modules (optional)

You can install the optional python dependencies with

pip install image jsonrpclib-pelix langid rpy2 simplejson textstat wordcloud

Standard DLATK functions can be run without these modules.

Install the IBM Wordcloud jar file (optional)

The IBM wordcloud module is our default. To install this you must sign up for a IBM DeveloperWorks account and download ibm-word-cloud.jar. Place this file into ../dlatk/lib/.

If you are unable to install this jar then you can use the python wordcloud module:

  1. pip install wordcloud
  2. Change wordcloud_algorithm='ibm' in ../dlatk/lib/ to wordcloud_algorithm='amueller'.

Note: You must install either the IBM Wordcloud jar or the Python wordcloud module to print wordclouds.

Mallet (optional)

Mallet can be used with DLATK to create LDA topics (see the Mallet LDA Interface tutorial). Directions on downloading and installing can be found here.

Command Line Interface

DLATK is run using which is added to /usr/local/bin during the installation process.

MySQL Configuration

  1. DLATK is highly dependent on MySQL. You must have this installed.
  2. Any calls to will open MySQL. We assume any table with text data has the following columns:
  • message: text data
  • message_id: unique numeric identifier for each message
  1. All lexicon tables are assumed to be in a database called permaLexicon (a sample database with this name is distributed with the release). To change this you must edit DEF_LEXICON_DB = 'permaLexicon'

Full List of Dependencies


  • matplotlib (>=1.3.1)
  • mysqlclient
  • nltk (>=3.1)
  • numpy
  • pandas (>=0.17.1)
  • python-dateutil (>=2.5.0)
  • scikit-learn (>=0.17.1)
  • scipy
  • SQLAlchemy (>=0.9.9)
  • statsmodels (>=0.6.1)


  • Stanford Parser
  • Tweet NLP

Python (optional)

  • image
  • jsonrpclib-pelix (>=0.2.8)
  • langid (>=1.1.4)
  • rpy2 (2.6.0)
  • simplejson (>=3.3.1)
  • textstat (>=0.6.1)
  • wordcloud (>=1.1.3)

Other (optional)

  • IBM Wordcloud (for wordcloud visualization)
  • Mallet (for creating LDA topics)

Install Issues

See Install FAQs for more info.