zad47.py

import pandas as pd
import mmh3
from sklearn.cluster import KMeans
import numpy as np

f = open("Szekspir/RomeoJuliet.txt")
file_content = f.read()
print(file_content)

scenes = file_content.split("Scene ")[1:]

min_hash = [[] for _ in range(len(scenes))]

for id, scene in enumerate(scenes):
    tokens = scene.split() 
    if len(tokens) < 3:
        shingles = [scene]
    else:
        shingles = [' '.join(tokens[i:i+3]) for i in range(len(tokens) - 2)]
        
    for it in range(100):
        min_hash_value = min(mmh3.hash(shingle, it) for shingle in shingles)
        min_hash[id].append(min_hash_value)

similarity_scores = [[0] * len(scenes) for _ in range(len(scenes))]

for i in range(len(scenes)):
    for j in range(i + 1, len(scenes)):
        similarity_scores[i][j] = sum(1 for a, b in zip(min_hash[i], min_hash[j]) if a == b) / 100
        print(f"Similarity between scene {i} and scene {j}: {similarity_scores[i][j]:.4f}")
        
max_value = max(max(row) for row in similarity_scores)
min_value = min(min(row) for row in similarity_scores if max(row) != 0)

print(f"Max similarity: {max_value}")
print(f"Min similarity: {min_value}")    

X = np.array(similarity_scores)

num_clusters = 3 
kmeans = KMeans(n_clusters=num_clusters, random_state=42, n_init=10)

kmeans.fit(1-X)
labels = kmeans.labels_

clusters = {i: [] for i in range(num_clusters)}

for scene_idx, cluster_id in enumerate(labels):
    clusters[cluster_id].append(scene_idx)

for cluster_id, scene_indices in clusters.items():
    print(f"Cluster {cluster_id}")
    print(f"  Indices: {scene_indices}")
    print("-" * 20)