You can deploy this model and start earning money today!

ludyhasby/
hoax_sahih_AI

Name: hoax_sahih_AI
Availability: InStock

text-classification

apache-2.0

Views

No views yet

Hoax Detection SahihAI 📰🤖

Sahih AI

Proyek ini bertujuan untuk mendeteksi apakah sebuah berita kemungkinan benar atau hoax menggunakan model deep learning berbasis TensorFlow. Model dan tokenizer di-host di 🤗 Hugging Face Hub.

Fitur

Preprocessing teks (normalisasi slang, penghapusan stopwords, pembersihan tanda baca).
Konversi emoji menjadi representasi kata.
Tokenisasi & padding otomatis sebelum inference.
Model klasifikasi biner (0 = Berita Benar, 1 = Berita Hoax).

Instalasi

Clone model:

bash

git clone https://huggingface.co/ludyhasby/hoax_sahih_AI
cd hoax_sahih_AI
pip install -r requirements.txt

Inference

Berikut adalah contoh penggunaan model

1. Import Library yang diperlukan, packages statics sudah kami masukkan kedalam direktori

python

import pickle
import tensorflow as tf
from tensorflow.keras.preprocessing.sequence import pad_sequences
from huggingface_hub import hf_hub_download
from statics import slang_dict2, stop_words, emoji_dict, max_length
import re

2. Deklarasikan beberapa fungsi penting

python

def load_important(token_hf):
    # --- Step 1. Load Model ---
    model_path = hf_hub_download(
        repo_id="ludyhasby/hoax_sahih_AI",
        filename="hoax_detection.h5", 
        token=token_hf   # masukkan token Anda
    )
    model = tf.keras.models.load_model(model_path, compile=False)

    # --- Step 2. Load Tokenizer ---
    tokenizer_path = hf_hub_download(
        repo_id="ludyhasby/hoax_sahih_AI",
        filename="tokenizer_hoax.pickle"
    )
    with open(tokenizer_path, "rb") as handle:
        tokenizer = pickle.load(handle)
    return model, tokenizer

# --- Step 3. Preprocessing Things ---
def replace_emot(teks, emoji_dict):
    for j, emoticon in (emoji_dict["Emoji"].items()):
        tag = emoji_dict["tag_indo"][j]
        teks = teks.replace(emoticon, f" {tag}")
    return teks

def normalize_slang(text, slang_dict):
  list_word =  text.split()
  for word in list_word:
    for j, original in slang_dict['original'].items():
      if word == original:
        text = text.replace(word, slang_dict['replacement'][j])
  return text

def teks_to_pad(teks):
    teks_seq = tokenizer.texts_to_sequences(teks)
    teks_pad = pad_sequences(teks_seq, maxlen=max_length, truncating=trunc_type, padding=pad_type)
    return teks_pad

def preprocessing(text, emot_f, slang_dict, STOP_PREP):
    text = replace_emot(text, emot_f)
    text = re.sub(r'[^\w\s]', ' ', text)
    text = text.lower()
    text = re.sub(r'username', '', text)
    text = normalize_slang(text, slang_dict)
    print(text)
    # Mengubah string menjadi list kata
    words = text.split()
    filtered_words = [word for word in words if word.lower() not in STOP_PREP]
    text = ' '.join(filtered_words)
    text = re.sub(r"\d+", "", text)
    text = re.sub(r'[ ]+', ' ', text)

    # Tokenizer
    padded = teks_to_pad([text])
    return padded

def decode_label(encode):
  if encode == 0:
    return "Berita Kemungkinan Benar"
  elif encode == 1:
    return "Berita Kemungkinan Salah [Hoax]"
  return "Out of Bound !"

def main_inference(teks, emoji_dict, slang_dict, STOP_PREP):
    padnya = preprocessing(teks, emoji_dict, slang_dict2, STOP_PREP)
    pred = model.predict(padnya)
    bin_result = (pred >= 0.5).astype(int)
    probability = pred[0][0] if bin_result==1 else 1-pred[0][0]
    print(decode_label(bin_result))
    print(probability)

3. Contoh Penggunaan

python

berita = input("Silahkan masukkan berita Anda: ")
main_inference(berita, emoji_dict, slang_dict2, stop_words)

4. Struktur Model

.
├── hoax_detection.h5      # Model terlatih
├── tokenizer_hoax.pickle  # Tokenizer
├── hoax_inference.py      # Script Contoh Penggunaan Siap Gunakan (Lokal)
├── requirements.txt
├── statics.py             # Modul Statis yang diperlukan saat proses inferensia
├── logo_sahihAI.png      
└── README.md