SoDehghan's picture
Update models/hsd_tr.py
48a6951 verified
raw
history blame
2.33 kB
import streamlit as st
from transformers import pipeline
# Turkish
#sentiment_pipeline_tr = pipeline(task = "text-classification", model = "SoDehghan/BERTurk-hate-speech-detection") # "gritli/bert-sentiment-analyses-imdb"
#sentiment_pipeline_tr_test = pipeline(task = "text-classification", model = "SoDehghan/test")
strength_pipeline_tr = pipeline(task = "text-classification", model = "SoDehghan/BERTurk-hate-speech-strength-prediction")
def write():
st.markdown(
"""
# Hate Speech Detection in Turkish
"""
)
tr_input = st.text_area("Enter your text here:", height=50, key="tr_input") #height=30
if st.button("Model prediction", key="tr_predict"):
st.write(" ")
with st.spinner('Generating predictions...'):
#result_sentiment_tr = sentiment_pipeline_tr(tr_input)
#sentiment_tr = result_sentiment_tr[0]["label"]
#label_dict_sentiment = {'LABEL_1': 'Detection: Hate ❌', 'LABEL_0': 'Detection: Non-hate βœ…'}
#sentiment_tr = label_dict_sentiment[sentiment_tr]
#result_sentiment_tr_test = sentiment_pipeline_tr_test(tr_input)
#sentiment_tr_test = result_sentiment_tr_test[0]["label"]
#label_dict_sentiment = {'LABEL_1': 'Detection: Hate ❌', 'LABEL_0': 'Detection: Non-hate βœ…'}
#sentiment_tr_test = label_dict_sentiment[sentiment_tr_test]
result_strength_tr = strength_pipeline_tr(tr_input)
strength_tr = result_strength_tr[0]["label"]
label_dict_strength = {'LABEL_0': 'Strength: 0 (No-hate)', 'LABEL_1': 'Strength: 1 (Insult)', 'LABEL_2': 'Strength: 2 (Exclusion)',
'LABEL_3': 'Strength: 3 (Wishing harm)', 'LABEL_4': 'Strength: 4 (Threatening harm)'}
label_dict_sentiment = {'LABEL_0': 'Detection: No-hate βœ…', 'LABEL_1': 'Detection: Hate ❌', 'LABEL_2': 'Detection: Hate ❌',
'LABEL_3': 'Detection: Hate ❌', 'LABEL_4': 'Detection: Hate ❌',}
sentiment_tr = label_dict_sentiment[strength_tr]
strength_tr = label_dict_strength[strength_tr]
st.write(sentiment_tr)
st.write(strength_tr)
#st.write(sentiment_tr_test)
#st.success(sentiment_tr)
#st.success(strength_tr)