train_predict.yml

# Environments
project_env: './env'

# Directories

## Querying
query_out_dir: 'out/original_query'
last_date_dir: 'out/last_query_date'

## Classification
classif_splits_dir: 'data/classif_splits'
classif_train_outdir: 'out/classif_train_out'
classif_benchmark_dir: 'out/benchmarks/classif'
classif_log_dir: 'out/logs/classif'
classif_out_dir: 'out/original_query/classification'

## NER
ner_splits_dir: 'data/ner_splits'
ner_train_outdir: 'out/ner_train_out'
ner_benchmark_dir: 'out/benchmarks/ner'
ner_log_dir: 'out/logs/ner'
ner_out_dir: 'out/original_query/ner'

## URL Extraction
extract_url_dir: 'out/original_query/url_extraction'

## Name processing
processed_names_dir: 'out/original_query/processed_names'

## Initial deduplication
initial_dedupe_dir: 'out/original_query/initial_deduplication'

## For manual review
for_manual_review_dir: 'out/original_query/for_manual_review'

## Manually reviewed
manually_reviewed_dir: 'out/original_query/manually_reviewed'

## Processed manual review
processed_manual_review: 'out/original_query/processed_manual_review'

## URL Checking
check_url_dir: 'out/original_query/url_checking'

## Additional metadata from EuropePMC
epmc_meta_dir: 'out/original_query/epmc_meta'

## Processed country codes
processed_countries: 'out/original_query/processed_countries'

## Data analysis
analysis_dir: 'analysis/'
figures_dir: 'analysis/figures'

# Input files
classif_data: 'data/manual_classifications.csv'
ner_data: 'data/manual_ner_extraction.csv'

# File with configuration settings for the models
models: 'config/models_info.tsv'

# Dates used for initial query
initial_query_start: 2011
initial_query_end: 2021
query_string: 'config/query.txt'

# Ratios used for data splitting
split_ratios: '0.7 0.15 0.15'

# Metrics used for choosing best model/epoch
class_criteria_metric: 'precision'
ner_criteria_metric: 'f1'

# Number of epochs
classif_epochs: 10
ner_epochs: 10

# Filtering parameters
max_urls: 2
min_best_name_prob: 0.978

# URL checking
chunk_size: 200
num_tries: 3
backoff: 0.5

# Getting metadata from EuropePMC
epmc_chunk_size: 20

# Processng country names
country_format: 'full'

# Input files for data analysis
## Data analysis can either be run on the newly generated output files
## or on the files stored in the repository. Comment/uncomment below
## to choose which files to use.

## Newly generated files
classification_train_stats: 'out/classif_train_out/combined_train_stats/combined_stats.csv'
classification_test_stats: 'out/classif_train_out/combined_test_stats/combined_stats.csv'
ner_train_stats: 'out/ner_train_out/combined_train_stats/combined_stats.csv'
ner_test_stats: 'out/ner_train_out/combined_test_stats/combined_stats.csv'
final_inventory_file: 'out/original_query/processed_countries/predictions.csv'

## Stored files
# classification_train_stats: 'data/classif_metrics/combined_train_stats.csv'
# classification_test_stats: 'data/classif_metrics/combined_test_stats.csv'
# ner_train_stats: 'data/ner_metrics/combined_train_stats.csv'
# ner_test_stats: 'data/ner_metrics/combined_test_stats.csv'
# final_inventory_file: 'data/final_inventory_2022.csv'

# Credentials file for FAIRsharing
fair_login_file: "config/fairsharing_login.json"

# Manually curated funding agency countries
curated_funders: "analysis/funders_geo_200.csv"