Getting Started

This document will show you how to get up and running with Fonduer. We’ll show you how to get everything installed and your machine so that you can walk through real examples by checking out our Tutorials.

Installing Non-Python Dependencies

Fonduer relies on a couple of non-Python applications. You’ll need to install these and be sure are on your PATH.

For OS X using homebrew:

$ brew install poppler
$ brew install postgresql@10
$ brew install libpng freetype pkg-config
$ brew install libomp #
$ brew install imagemagick

On Debian-based distros:

$ sudo apt update
$ sudo apt install libxml2-dev libxslt-dev python3-dev libpq-dev
$ sudo apt build-dep python-matplotlib
$ sudo apt install poppler-utils
$ sudo apt install postgresql
$ sudo apt install libmagickwand-dev


Fonduer requires PostgreSQL version 9.6 or higher.


Fonduer requires poppler-utils to be version 0.36.0 or later. Otherwise, the -bbox-layout option is not available for pdftotext (see changelog). It is recommended to use poppler-utils version 0.48.0 or later to avoid a known bug.


Use Wand (>=0.5.0) with ImageMagick7 as Wand (<0.5.0) does not support ImageMagick7.

Installing the Fonduer Package

Then, install Fonduer by running:

$ pip install fonduer


Fonduer only supports Python 3. Python 2 is not supported.


For the Python dependencies, we recommend using a virtualenv, which will allow you to install Fonduer and its python dependencies in an isolated Python environment. Once you have virtualenv installed, you can create a Python 3 virtual environment as follows.:

$ virtualenv -p python3.6 .venv

Once the virtual environment is created, activate it by running:

$ source .venv/bin/activate

Any Python libraries installed will now be contained within this virtual environment. To deactivate the environment, simply run:

$ deactivate

Downloading spaCy language models

Language models introduced recently cannot be downloaded by Fonduer. Those models should be downloaded and their shortcuts should be created as below:

$ python -m spacy download ja_core_news_sm
$ python -m spacy link ja_core_news_sm ja
$ python -m spacy download zh_core_web_sm
$ python -m spacy link zh_core_web_sm zh

The Fonduer Pipeline

The Fonduer pipeline can be broken into five phases.

  1. Parsing

    In this first stage, an input corpus of richly formatted documents is parsed into Fonduer’s data model.

  2. Mention and Candidate Extraction

    Here, we initialize the knowledge base with the user’s target schema. Users define Mentions using Matchers, and then combine Mentions to create Candidates. Throttlers can also (optionally) be added to filter out invalid Candidates to achieve better class balance.

  3. Multimodal Featurization

    Fonduer then featurizes each candidate with features from multiple modalities.

  4. Supervision

    Next, users provide labeling functions (which can leverage our data model utilities) to provide weak supervision.

  5. Classification

    Finally, Fonduer provides machine learning models which are used to classify each Candidate.

To demonstrate how to set up and use Fonduer in your applications, we walk through each of these phases in real-world examples in our Tutorials.

Check out the Fonduer paper for more details about the system.