configuration-sample.yml

---
## Prediction
output_path: ./prediction/configuration-sample

## Probase
lemma_dir: ./spacy-probase/spacy/
model_dir: ./model/configuration-sample/

## Data
train: ./data/training/train.txt
dev: ./data/training/validation.txt
test: ./ref/truth.txt

## Features
data_type:
    - train
    - dev
    - test

pretrain_embeddings:
    - name: GloVe
      path: ./data/vectors/glove.6B.300d.w2v
      is_bin: False

    - name: Word2Vec
      path: ./data/vectors/GoogleNews-vectors-negative300.bin
      is_bin: True

    - name: FastText
      path: ./data/vectors/wiki.en.vec
      is_bin: False

## Evaluation
random_seed: 1

# Which data to use for cross validation and number of training folds
# uses is a list of data_type, i.e. train/dev/test
# fold is number of folds for cv
cross_validation_configs:
    - uses:
        - dev
      folds: 5

# How to split the data for training and testing
# train and test are a list of data_type, i.e. train/dev/test
#
# For instance:
#
#     - train:
#        - train
#        - dev
#      test:
#        - test
#
# means to train with the train and dev partition together,
# and test with test partition
split_configs:
    - train:
        - dev
      test:
        - test

    - train:
        - train
        - dev
      test:
        - test

    - train:
        - train
      test:
        - test

    - train:
        - train
      test:
        - dev

# What feature to use in each training process
# For instance:
#
# - - GloVe
#   - Probase
#
# means to use features from GloVe and Probase together (concatenate)
# and
#
# - - Probase
#
# means to use only use features from Probase
feature_configs:
    - - Probase

    - - GloVe

    - - GloVe
      - Probase

    - - Word2Vec

    - - Word2Vec
      - Probase

    - - FastText

    - - FastText
      - Probase

# Which classifier to use for training?
classifier_configs:
    - logistic
    - dtc
    - svm-rbf
    - rf
    - LinearSVC