run_wikinewsmax.sh

#!/bin/bash

# Get the parent directory of this script
SCRIPT_DIR="$(dirname $(realpath "$0"))"

# List of genres to be evaluated
DATASETS=(
    "dediac_max_gold"
    "dediac_orig_gold"
)

DATA_DIR="${SCRIPT_DIR}/data/wikinewsmax"
PRED_DIR="${SCRIPT_DIR}/output/predictions/wikinewsmax"
EVAL_DIR="${SCRIPT_DIR}/output/eval/wikinewsmax"
DB_ORIG_PATH="${SCRIPT_DIR}/databases/calima-msa-s31.db"
DB_EXT_PATH="${SCRIPT_DIR}/databases/calima-msa-s31-extended.db"

# Function that generates diacritization predictions using original CAMeL Tools
# for a given dataset
function gen_predictions_original () {
    DATASET="$1"
    INPUT_PATH="${DATA_DIR}/${DATASET}.tsv"
    OUTPUT_PATH="${PRED_DIR}/${DATASET}.original.tsv"
    DB_PATH="${SCRIPT_DIR}/databases/calima-msa-s31.db"

    printf ">>> Generating diac predictions for ${DATASET} using original CAMeL Tools...\n"

    python3 "${SCRIPT_DIR}/gen_predictions_sent.py" "${INPUT_PATH}" \
        --db "${DB_ORIG_PATH}" \
        --ct-noctx \
        --output "${OUTPUT_PATH}"

    if [ $? -ne 0 ]; then
        exit 1
    fi
}

# Function that generates diacritization predictions using modified CAMeL Tools
# with extentions for a given dataset
function gen_predictions_extended () {
    DATASET="$1"
    INPUT_PATH="${SCRIPT_DIR}/data/wikinewsmax/${DATASET}.tsv"
    OUTPUT_PATH="${SCRIPT_DIR}/output/predictions/wikinewsmax/${DATASET}.extended.tsv"
    DB_PATH="${SCRIPT_DIR}/databases/calima-msa-s31-extended.db"

    printf ">>> Generating diac predictions for ${DATASET} using extended CAMeL Tools...\n"

    python3 "${SCRIPT_DIR}/gen_predictions_sent.py" "${INPUT_PATH}" \
        --db "${DB_EXT_PATH}" \
        --ctpp-fullctx \
        --output "${OUTPUT_PATH}"

    if [ $? -ne 0 ]; then
        exit 1
    fi
}

# Function that evaluates predictions generated using original CAMeL Tools for
# a given list of genres
function evaluate_original () {
    DATASETS="$1"
    INPUT_DIR="${SCRIPT_DIR}/data/wikinewsmax"

    for DATASET in ${DATASETS[@]}; do
        OUTPUT_PATH="${EVAL_DIR}/${DATASET}.original.tsv"
        DATASET_PATH_ARG="${DATASET}:${PRED_DIR}/${DATASET}.original.tsv"

        printf ">>> Evaluating predictions for ${DATASET} using original CAMeL Tools...\n"

        python3 "${SCRIPT_DIR}/evaluate.py" --output "${OUTPUT_PATH}" \
            --ct-noctx --oov \
            --norm-diac --norm-consonants --use-gold-alt \
            "${DATASET_PATH_ARG}"

        if [ $? -ne 0 ]; then
            exit 1
        fi
    done
}

# Function that evaluates predictions generated using modified CAMeL Tools with
# extentions for a given list of genres
function evaluate_extended () {
    DATASETS="$1"
    INPUT_DIR="${SCRIPT_DIR}/data/wikinewsmax"

    for DATASET in ${DATASETS[@]}; do
        OUTPUT_PATH="${EVAL_DIR}/${DATASET}.extended.tsv"
        DATASET_PATH_ARG="${DATASET}:${PRED_DIR}/${DATASET}.extended.tsv"

        printf ">>> Evaluating predictions for ${DATASET} using extended CAMeL Tools...\n"

        python3 "${SCRIPT_DIR}/evaluate.py" --output "${OUTPUT_PATH}" \
            --ctpp-fullctx --oov \
            --norm-diac --norm-consonants --use-gold-alt \
            "${DATASET_PATH_ARG}"

        if [ $? -ne 0 ]; then
            exit 1
        fi
    done
}

# Use custom CAMeL Tools data directory
export CAMELTOOLS_DATA="${SCRIPT_DIR}/envs/ct_data"

# Create output directories if they don't already exist
mkdir -p "${SCRIPT_DIR}/output/predictions/wikinewsmax"
mkdir -p "${SCRIPT_DIR}/output/eval/wikinewsmax"

# Deactivate the current environment (if any)
deactivate &> /dev/null

# Activate the modified CAMeL Tools environment
source "${SCRIPT_DIR}/envs/modified/bin/activate"

# Generate predictions using modified CAMeL Tools with extentions
for DATASET in ${DATASETS[@]}; do
    gen_predictions_extended ${DATASET}
done

# Deactivate the modified CAMeL Tools environment
deactivate

# Activate the original CAMeL Tools environment
source "${SCRIPT_DIR}/envs/original/bin/activate"

# Generate predictions using original CAMeL Tools
for DATASET in ${DATASETS[@]}; do
    gen_predictions_original ${DATASET}
done

# Evaluate original and extended predictions seperately
evaluate_original ${DATASETS}
evaluate_extended ${DATASETS}

# Deactivate the original CAMeL Tools environment
deactivate