4

detect.py

@@ -0,0 +1,35 @@
+from model import m
+from preprocessor import frs
+import pickle
+import time
+
+from sys import exit
+
+from tensorflow.keras.utils import pad_sequences as kps
+
+m.load_weights("model2.keras")
+tk = pickle.load(open('tokenizer.pickle', 'rb'))
+
+while True:
+    t = ""
+    try:
+        t = frs(input("Comment: "))
+    except EOFError:
+        print("\nExiting")
+        exit(0)
+
+    print(f"Processed: {t}")
+
+    s = tk.texts_to_sequences([t])
+    print(f"Sequence: {s[0]}")
+
+    ps = kps(s, maxlen = 100)
+
+    p = m.predict(ps)
+
+    if p >= 0.75:
+        print(f"Result: positive ({p[0]})\n")
+    elif p <= 0.25:
+        print(f"Result: negative ({p[0]})\n")
+    else:
+        print(f"Result: unsure ({p[0]})\n")

main.py

@@ -10,6 +10,9 @@ from tensorflow.keras.utils import pad_sequences as kps
 import numpy as np
 import pandas as pd
 
+import pickle
+
+
 print("I")
 t = pd.read_csv("prepped_train.csv",
                 header = None,
@@ -22,6 +25,9 @@ r = t['r'].astype(str)
 tk = kT(num_words = 6000)
 tk.fit_on_texts(r)
 
+with open('tokenizer.pickle', 'wb') as f:
+    pickle.dump(tk, f, protocol = pickle.HIGHEST_PROTOCOL)
+
 print("F")
 
 s = tk.texts_to_sequences(r)
@@ -50,6 +56,7 @@ m.compile(optimizer = ko.Lion(learning_rate = 0.0005),
           metrics = ['accuracy'])
 '''
 
+'''
 from model import m
 
 ckpt = kc.ModelCheckpoint('model2.keras',
@@ -64,3 +71,4 @@ history = m.fit(ts,
                 batch_size = 1024,
                 validation_split = 0.1,
                 callbacks = [ckpt])
+'''

preprocessor.py

@@ -5,9 +5,9 @@ from nltk.tokenize import word_tokenize
 from nltk.stem import WordNetLemmatizer
 from spellchecker import SpellChecker as sc
 
-nltk.download("stopwords")
+#nltk.download("stopwords")
-nltk.download("punkt_tab")
+#nltk.download("punkt_tab")
-nltk.download("wordnet")
+#nltk.download("wordnet")
 
 def fr(r):
     r = r.lower()