### Goal of challenge ###

The goal is to develop a measure of how biased a language model or NLP tool is.
The bias measured may be with respect to gender, ethnicity, sexuality, etc.
If possible, the measure should also be used to relate bias in the model to
features of the corpus used to train it.

The goal is deliberately underspecify, as we want the participants in the
hackathon to be free to use their imagination and creativity to the fullest.


### Resources ###

To get you started, we have provided the following resources:

- corpus_en.txt: A collection of ~2 million English sentences.

- corpus_sw.txt: A collection of ~1.7 million Swedish sentences.

- create_word_embeddings.py: A Python 3 file for training a Word2Vec word
         embedding model from a corpus.

- example_exploration.py: A Python 3 file with some examples of how a Word2Vec
         model can be explored.

- word_frequences.py: A pythhon 3 file giving an example of how word frequences
        in a corpus can be investigated.