calculate utterance similarity using DTW algorithm.

dtw.py

from scipy.spatial.distance import euclidean
from fastdtw import fastdtw
import numpy as np

def calculate_similarity_percentage(distance, max_distance):
    similarity_percentage = (1 - distance / max_distance) * 100
    return similarity_percentage

def main():
    test_sets = [
        ("Not yet. Oh, I can see it. Sorry, I can see it.", 
         "Not yet, Oh, I can see it. Sorry I can see."),
        ("test3", 
         "test4"),
         ("1234567890", 
         "0123456789"),
         ("100%, perfectly matched", 
         "100%, perfectly matched"),
         ("100%, perfectly matched", 
         "100%, perfectly matche."),
         ("non sementic  parts is modified, so it's not  p roblem.", 
         "non-sementic parts is modified. so it's not a problem."),
         ("I have a lots of experience in NLP. But NLP means Non-Lexical-Problem.", 
         "I have a lots of experience in NLP."),
        # Add more test sets as needed
    ]

    for text1, text2 in test_sets:
        
        ascii_values1 = [ord(char) for char in text1]
        ascii_values2 = [ord(char) for char in text2]

        # Reshape for DTW
        ascii_values1 = np.array(ascii_values1).reshape(-1, 1)
        ascii_values2 = np.array(ascii_values2).reshape(-1, 1)

        # Compute the DTW distance
        distance, path = fastdtw(ascii_values1, ascii_values2, dist=euclidean)
        # distance = calculate_dtw_distance(text1, text2)
        max_distance = max(len(text1), len(text2)) * (122 - 32)
        similarity_percentage = (1 - distance / max_distance) * 100
        # similarity_percentage = calculate_similarity_percentage(distance, max_distance)
        print("-------------------------")
        print(f"Text 1: {text1}")
        print(f"Text 2: {text2}")
        # print(f"DTW Distance: {distance}")
        print(f"Similarity Percentage: {similarity_percentage:.2f}%")
        print("-------------------------")

# if __name__ == "__main__":
main()