Model description

This model is a fine-tuned model of intfloat/multilingual-e5-large, trained with Indonesian police news data.

How to use this model:

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("faizaulia/e5-fine-tune-polri-news-emotion")
model = AutoModelForSequenceClassification.from_pretrained("faizaulia/e5-fine-tune-polri-news-emotion")

Label description:

0: Angry, 1: Fear, 2: Sad, 3: Neutral, 4: Happy, 5: Love

Input text example:

LAMPUNG, KOMPAS.com - Komplotan perampok yang menyekap satu keluarga di Kabupaten Lampung Timur ditembak aparat kepolisian. Komplotan ini menggondol uang sebanyak Rp 50 juta milik korban. Kapolres Lampung Timur, AKBP M Rizal Muchtar mengatakan, tiga dari empat pelaku ini telah ditangkap pada Senin (27/2/2023) dini hari.

Preprocesssing:

nltk.download('stopwords')
nltk.download('wordnet')

stop_words = set(stopwords.words('indonesian'))

def remove_stopwords(text):
    words = text.split()
    words = [word for word in words if word not in stop_words]
    return ' '.join(words)

def clean_texts(text):
    text = re.sub('\n',' ',text) # Remove every '\n'
    text = re.sub('  +', ' ', text) # Remove extra spaces
    text = re.sub('[\u2013\u2014]', '-', text) # Sub — and – char to -
    text = re.sub('(.{0,40})-', '', text) # Remove news website/location at the beginning
    text = re.sub(r'[^a-zA-Z\s]', '', text) # Remove non alphanbet characters
    return text

def preprocess_text(text):
    text = text.lower()
    text = clean_texts(text)
    text = remove_stopwords(text)
    return text