word generation integration with backend

App/Backend/server.py

@@ -2,6 +2,7 @@ import numpy as np
 from flask import Flask, request, jsonify, request
 import pickle
 from word_card_game import wordGameData
+from word_generation import get_similar_words
 
 
 app = Flask(__name__)
@@ -12,6 +13,7 @@ app = Flask(__name__)
 
 
 # send a json {'exp':1.8,} as a post request to make a prediction
+'''
 @app.route('/api/predict',methods=['POST'])
 def predict():
     data = request.get_json(force=True)
@@ -24,7 +26,28 @@ def predict():
 def default_get():
     return "<p>HereMe Backend !</p>"
 
+'''
+@app.route('/api/word-game', methods=['GET'])
+def word_game_api():
+    w1 = request.args.get('w1')
+    w2 = request.args.get('w2')
+    w3 = request.args.get('w3')
 
+    if not all([w1, w2, w3]):
+        return jsonify({'error': 'All three words must be provided'}), 400
+
+    data = wordGameData(w1, w2, w3)
+    return jsonify(data)
+
+@app.route('/api/similar-words', methods=['GET'])
+def similar_words_api():
+    word = request.args.get('word')
+
+    if not word:
+        return jsonify({'error': 'A word must be provided'}), 400
+
+    similar_words = get_similar_words(word)
+    return jsonify({'similar_words': similar_words})
 
 if __name__ == '__main__':
     app.run(port=5000, debug=True)
\ No newline at end of file

App/Backend/word_generation.py

+import torch
+from transformers import RobertaTokenizer, RobertaForMaskedLM
+
+# Load the pretrained RoBERTa model and tokenizer
+tokenizer = RobertaTokenizer.from_pretrained('roberta-base')
+model = RobertaForMaskedLM.from_pretrained('roberta-base')
+
+def get_similar_words(input_word, top_k=3):
+    # Create a masked sentence with the input word
+    masked_sentence = f"The {input_word} is related to the {tokenizer.mask_token}."
+    
+    # Tokenize the masked sentence
+    inputs = tokenizer(masked_sentence, return_tensors='pt')
+    
+    # Get the index of the mask token
+    mask_token_index = torch.where(inputs['input_ids'][0] == tokenizer.mask_token_id)[0].item()
+    
+    # Predict words for the mask token
+    with torch.no_grad():
+        output = model(**inputs)
+        predictions = output.logits[0, mask_token_index]
+    
+    # Get the top k predicted words
+    top_k_indices = torch.topk(predictions, top_k).indices.tolist()
+    related_words = [tokenizer.decode(idx).strip() for idx in top_k_indices]
+    
+    return related_words