main.py

import data_processing
import ocr


def score(list_a, list_b):
    """Computes the score"""
    if not len(list_a) == len(list_b): raise Exception("Dimension mismatch")
    return sum(ea == eb for ea, eb in zip(list_a, list_b)) / len(list_a)

def main():
    # Load Data and split in test and train set
    words = data_processing.read_PA3Data()
    train_words = words[:80]
    test_words = words[80:]

    # Train logistig model and load picewise model
    logistig_model = data_processing.train_logreg_model(train_words)
    pairwise_model =  data_processing.read_PA3Models_pairwise()

    # Predict test words
    precition = [ocr.construct_network([l[0] for l in word], logistig_model, pairwise_model)
                 for word in test_words]

    # Calculate Scores
    print("Score for words: ", score(precition, [[l[1] for l in word] for word in test_words]))
    print("Score for letters: ", score([l for word in precition for l in word],
                                    [l[1] for word in test_words for l in word]))


##import numpy as np
##def gen_custom_pairwise(letters):
##    mod = np.zeros((26,26))
##    letter_pair = list(zip(letters[1:], letters[:-1]))
##    for i in range(26):
##        for j in range(26):
##            mod[i, j] = letter_pair.count(( chr(i + ord("a")), chr(j + ord("a")) ))
##    return mod +  0.1
##pairwise_model = gen_custom_pairwise([l[1] for word in test_words for l in word])

if __name__ == '__main__':
    main()