File size: 6,704 Bytes
cfc20e5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
import json
import re
from collections import defaultdict
import torch
import torch.nn as nn
import torch.optim as optim

def flatten_json_data(json_file_path):
    with open(json_file_path, 'r', encoding='utf-8') as f:
        data = json.load(f)

    flattened_data = []
    for dataset in data:
        for section in ["conversation_samples", "technical_terminology"]:
            if section in dataset:
                for category, samples in dataset[section].items():
                    for item in samples:
                        if "balochi" in item and "english" in item and "urdu" in item and "persian" in item: # Check if all keys are present
                            for balochi_sentence, english_sentence, urdu_sentence, persian_sentence in zip(item['balochi'], item['english'], item['urdu'], item['persian']):
                                flattened_data.append({
                                    "context": item.get('context', category),
                                    "balochi": balochi_sentence,
                                    "english": english_sentence,
                                    "urdu": urdu_sentence,
                                    "persian": persian_sentence
                                })
    return flattened_data

def parse_tsv_data(tsv_file_path):
    data = []
    with open(tsv_file_path, 'r', encoding='utf-8') as f:
        for line in f:
            balochi, english = line.strip().split('\t')
            data.append({"balochi": balochi, "english": english})
    return data

def create_vocab(tokenized_sentences):
    vocab = defaultdict(lambda: len(vocab))
    vocab['<PAD>'] = 0
    vocab['<UNK>'] = 1 # Add unknown token
    for sentence in tokenized_sentences:
        for word in sentence:
            vocab[word]
    return dict(vocab)

def tokenize_text(text):
    return re.findall(r"[\w']+(?:ء)?|[\u0600-\u06FF]+|[؟،۔٬؛٪٫٬]+|\S", text)

class SimpleRNNTranslator(nn.Module):
    def __init__(self, input_size, output_size, hidden_size):
        super(SimpleRNNTranslator, self).__init__()
        self.hidden_size = hidden_size
        self.embedding = nn.Embedding(input_size, hidden_size)
        self.rnn = nn.RNN(hidden_size, hidden_size, batch_first=True)
        self.fc = nn.Linear(hidden_size, output_size)

    def forward(self, x):
        embedded = self.embedding(x)  # Shape: [batch_size, seq_len, hidden_size]
        output, _ = self.rnn(embedded)  # Shape: [batch_size, seq_len, hidden_size]
        output = self.fc(output)  # Shape: [batch_size, seq_len, output_size]
        return output


def train_model(flattened_data, num_epochs=10, hidden_size=256, learning_rate=0.001):
    # Tokenize sentences
    balochi_sentences = [tokenize_text(entry['balochi']) for entry in flattened_data]
    english_sentences = [tokenize_text(entry['english']) for entry in flattened_data]

    # Create vocabularies
    balochi_vocab = create_vocab(balochi_sentences)
    english_vocab = create_vocab(english_sentences)

    input_size = len(balochi_vocab)
    output_size = len(english_vocab)

    # Use a unified max sequence length
    max_seq_len = max(
        max(len(sentence) for sentence in balochi_sentences),
        max(len(sentence) for sentence in english_sentences)
    )

    def encode_sentences(sentences, vocab, max_len):
        encoded = [
            [vocab.get(word, vocab['<UNK>']) for word in sentence] + [0] * (max_len - len(sentence))
            for sentence in sentences
        ]
        return torch.LongTensor(encoded)

    # Encode inputs and targets with the same max_seq_len
    X = encode_sentences(balochi_sentences, balochi_vocab, max_seq_len)  # Shape: [batch_size, max_seq_len]
    Y = encode_sentences(english_sentences, english_vocab, max_seq_len)  # Shape: [batch_size, max_seq_len]

    # Initialize model, loss, and optimizer
    model = SimpleRNNTranslator(input_size, output_size, hidden_size)
    criterion = nn.CrossEntropyLoss(ignore_index=0)  # Ignore padding token during loss computation
    optimizer = optim.Adam(model.parameters(), lr=learning_rate)

    # Training loop
    for epoch in range(num_epochs):
        model.train()
        optimizer.zero_grad()

        # Forward pass
        outputs = model(X)  # Shape: [batch_size, max_seq_len, output_size]
        outputs = outputs.view(-1, output_size)  # Flatten outputs for loss computation

        # Flatten target tensor
        target = Y.view(-1)  # Shape: [batch_size * max_seq_len]

        # Compute loss
        loss = criterion(outputs, target)

        # Backward pass and optimization
        loss.backward()
        optimizer.step()

        print(f"Epoch [{epoch + 1}/{num_epochs}], Loss: {loss.item():.4f}")

    return model, balochi_vocab, english_vocab, max_seq_len

def translate(model, balochi_sentence, balochi_vocab, english_vocab, max_balochi_len, max_output_len=20):
    tokenized_sentence = tokenize_text(balochi_sentence)
    encoded_sentence = [balochi_vocab.get(word, balochi_vocab['<UNK>']) for word in tokenized_sentence]
    padded_sentence = encoded_sentence + [0] * (max_balochi_len - len(encoded_sentence))
    input_tensor = torch.LongTensor([padded_sentence])

    # Initialize decoding
    model.eval()
    with torch.no_grad():
        output_tensor = model(input_tensor)  # Shape: [batch_size, seq_len, output_size]
        predicted_indices = torch.argmax(output_tensor, dim=2)[0]  # Take the first (and only) batch
        reverse_english_vocab = {idx: word for word, idx in english_vocab.items()}
        predicted_words = [reverse_english_vocab.get(idx.item(), '<UNK>') for idx in predicted_indices]
    
    return " ".join(predicted_words[:max_output_len])


# Example usage
json_file_path = "mergedv1.json"  # Replace with your JSON file path
tsv_file_path = "data_bal_en.tsv"  # Replace with your TSV file path

# Load and combine data
flattened_json_data = flatten_json_data(json_file_path)
flattened_tsv_data = parse_tsv_data(tsv_file_path)
flattened_data = flattened_json_data + flattened_tsv_data

# Check if data was loaded
if not flattened_data:
    print("Error: No data loaded from JSON or TSV. Check file paths and formats.")
else:
    model, balochi_vocab, english_vocab, max_balochi_len = train_model(flattened_data)

    balochi_input = "تئو چے کنغ پسند کنئے؟"
    translated_sentence = translate(model, balochi_input, balochi_vocab, english_vocab, max_balochi_len)
    print(f"Balochi Input: {balochi_input}")
    print(f"Translated Output: {translated_sentence}")