project.yml

title: "Polar Component"
description: |
  This example project shows how to implement a simple stateful component to
  score docs on semantic poles.

  The method here is based on SemAxis from [An et al
  2018](https://arxiv.org/abs/1806.05521). The basic idea is that given a set
  of word vectors and some seed poles, like "bad-good", it's possible to
  calculate reference vectors. The distance of document vectors from those
  reference vectors is like a sentiment or polar score of the document. While
  not as sophisticated as a trained model, it's easy to test with existing data.

  If you use enough poles, you can use the scores as semantic vectors that can
  make downstream tasks explainable. This is explored in the SemAxis paper as
  well as [Mathew et al 2020](https://arxiv.org/abs/2001.09876), "The Polar
  Framework". (Incorporating semantic vectors as features in a spaCy model is
  left as an exercise for the reader.) 

  **Note:** Because the data is hosted on Kaggle, it can't be automatically
  downloaded by `spacy project assets`, so you'll have to download it yourself.
  See [the assets section of this README](#assets) for the link.

vars:
  reviews: "assets/IMDB Dataset.csv"

# These are the directories that the project needs. The project CLI will make
# sure that they always exist.
directories: ["scripts", "assets"]

# Assets that should be downloaded or available in the directory. You can replace
# this with your own input data.
assets:
    - dest: ${vars.reviews}
      description: "IMDB Review Corpus. Download from [Kaggle](https://www.kaggle.com/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews)."

workflows:
  all:
    - evaluate

# Project commands, specified in a style similar to CI config files (e.g. Azure
# pipelines). The name is the command name that lets you trigger the command
# via "spacy project run [command] [path]". The help message is optional and
# shown when executing "spacy project run [optional command] [path] --help".
commands:
  - name: "evaluate"
    help: "Check output on sample data"
    script:
      - "python ./scripts/evaluate.py"
    deps:
      - ${vars.reviews}