import streamlit as st
import pandas as pd
import numpy as np
import re
import json
import joblib
from sklearn.feature_extraction.text import TfidfVectorizer
# Impor library tambahan
#import matplotlib.pyplot as plt
#import seaborn as sns
#import plotly.express as px
from wordcloud import WordCloud
import nltk
from nltk.corpus import stopwords
#from transformers import pipeline
# Fungsi untuk membersihkan teks dengan ekspresi reguler
#@st.cache_data
def clean_text(text):
# Tahap-1: Menghapus karakter non-ASCII
text = re.sub(r'[^\x00-\x7F]+', '', text)
# Tahap-2: Menghapus URL
text = re.sub(r'http[s]?://.[a-zA-Z0-9./_?=%+!]+', '', text)
text = re.sub(r'pic.twitter.com?.[a-zA-Z0-9./_?=%+!]+', '', text)
# Tahap-3: Menghapus mentions
text = re.sub(r'@[\w]+', '', text)
# Tahap-4: Menghapus hashtag
text = re.sub(r'#([\w]+)', '', text)
# Tahap-5 Menghapus 'amp' yang menempel pada '&' dan 'gt' yang menempel pada '&'
text = re.sub(r'&|>', '', text)
# Tahap-6: Menghapus karakter khusus (simbol)
text = re.sub(r'[!$%^&*@#()_+|~=`{}\[\]%\-:";\'<>?,./]', '', text)
# Tahap-7: Menghapus angka
text = re.sub(r'[0-9]+', '', text)
# Tahap-8: Menggabungkan spasi ganda menjadi satu spasi
text = re.sub(' +', ' ', text)
# Tahap-9: Menghapus spasi di awal dan akhir kalimat
text = text.strip()
# Tahap-10: Konversi teks ke huruf kecil
text = text.lower()
# Tahap-11: koreksi duplikasi tiga karakter beruntun atau lebih (contoh. yukkk)
# text = re.sub(r'([a-zA-Z])\1\1', '\\1', text)
#text = re.sub(r'(.)(\1{2,})', r'\1\1', text)
text = re.sub(r'(\w)\1{2,}', r'\1', text)
return text
@st.cache_resource
def load_file(kamus_path, kamus_sendiri_path):
# Membaca kamus kata gaul Salsabila
with open(kamus_path) as f:
data = f.read()
lookp_dict = json.loads(data)
# Dict kata gaul saya sendiri yang tidak masuk di dict Salsabila
with open(kamus_sendiri_path) as f:
kamus_sendiri = f.read()
kamus_gaul_baru = json.loads(kamus_sendiri)
# Menambahkan dict kata gaul baru ke kamus yang sudah ada
lookp_dict.update(kamus_gaul_baru)
nltk.download("stopwords")
stop_words = set(stopwords.words("indonesian"))
additional_stopwords = [] # Ganti dengan kata-kata yang ingin Anda tambahkan
stop_words.update(additional_stopwords)
# Hapus beberapa kata dari kamus stopwords agar tidak terhapus pada tweets
words_to_remove = ['lama', 'datang', 'sekarang', 'percuma', 'jauh', 'waktu', 'kurang', 'bagaimana', 'gimana','tanya','berapa','jadwal','info','naik' ]
for word in words_to_remove:
if word in stop_words:
stop_words.remove(word)
tfidf_vectorizer = joblib.load(tfidf_model_path)
model_ensemble = joblib.load('ensemble_clf_soft_smote.joblib')
#model_rf
model_nb = joblib.load('naive_bayes_model_smote.joblib')
model_lr = joblib.load('logreg_model_smote.joblib')
return lookp_dict, stop_words, tfidf_vectorizer, model_ensemble, model_nb, model_lr
# Fungsi untuk normalisasi kata gaul
#@st.cache_data
def normalize_slang(text, slang_dict):
words = text.split()
normalized_words = [slang_dict.get(word, word) for word in words]
return ' '.join(normalized_words)
#---------------------------------------------------NLTK Remove Stopwords----------------------------------------------------------------------
#@st.cache_data
def remove_stopwords(text, stop_words):
# Pecah teks menjadi kata-kata
words = text.split()
# Hapus stopwords bahasa Indonesia
words = [word for word in words if word not in stop_words]
return " ".join(words)
#---------------------------------------------------TFIDF----------------------------------------------------------------------
# Memuat model TF-IDF dengan joblib (pastikan path-nya benar)
# Fungsi untuk ekstraksi fitur TF-IDF
#@st.cache_data
#def extract_tfidf_features(texts, _tfidf_vectorizer):
# tfidf_matrix = tfidf_vectorizer.transform(texts)
# return tfidf_matrix
#---------------------------------------------------Milih Model----------------------------------------------------------------------
# Fungsi untuk memilih model berdasarkan pilihan pengguna
def select_sentiment_model(selected_model, model_enesmble, model_nb, model_lr):
if selected_model == "Ensemble":
model = model_ensemble
elif selected_model == "Random Forest":
model = model_ensemble
elif selected_model == "Naive Bayes":
model = model_nb
elif selected_model == "Logistic Regression":
model = model_lr
else:
# Fallback ke model default jika pilihan tidak valid
model = model_ensemble
return model
# Fungsi untuk prediksi sentimen
# def predict_sentiment(text, _sentiment_model, _tfidf_vectorizer, slang_dict):
# # Tahap-1: Membersihkan dan normalisasi teks
# cleaned_text = clean_text(text)
# norm_slang_text = normalize_slang(cleaned_text, slang_dict)
# # Tahap-2: Ekstraksi fitur TF-IDF
# tfidf_matrix = _tfidf_vectorizer.transform([norm_slang_text])
# # Tahap-3: Lakukan prediksi sentimen
# sentiment = _sentiment_model.predict(tfidf_matrix)
# # Tahap-4: Menggantikan indeks dengan label sentimen
# labels = {0: "Negatif", 1: "Netral", 2: "Positif"}
# sentiment_label = labels[int(sentiment)]
# if sentiment == "Positif":
# emoticon = "đ" # Emotikon untuk sentimen positif
# elif sentiment == "Negatif":
# emoticon = "đ" # Emotikon untuk sentimen negatif
# else:
# emoticon = "đ" # Emotikon untuk sentimen netral
# return sentiment_label, emoticon
def predict_sentiment(text, _sentiment_model, _tfidf_vectorizer, slang_dict):
# Tahap-1: Membersihkan dan normalisasi teks
cleaned_text = clean_text(text)
norm_slang_text = normalize_slang(cleaned_text, slang_dict)
# Tahap-2: Ekstraksi fitur TF-IDF
tfidf_matrix = _tfidf_vectorizer.transform([norm_slang_text])
# Tahap-3: Lakukan prediksi sentimen
sentiment = _sentiment_model.predict(tfidf_matrix)[0] # Ambil elemen pertama dari hasil prediksi
# Tahap-4: Menggantikan indeks dengan label sentimen
labels = {0: "Negatif", 1: "Netral", 2: "Positif"}
sentiment_label = labels[sentiment]
# Tahap-5: Tentukan emoticon berdasarkan label sentimen
emoticons = {"Negatif": "đ", "Netral": "đ", "Positif": "đ"}
emoticon = emoticons.get(sentiment_label, "đ") # Default emoticon untuk label tidak dikenal
return sentiment_label, emoticon
@st.cache_data
def buat_chart(df, target_year):
target_year = int(target_year)
st.write(f"Bar Chart Tahun {target_year}:")
# Ambil bulan
df['Date'] = pd.to_datetime(df['Date']) # Convert 'Date' column to datetime
df['month'] = df['Date'].dt.month
df['year'] = df['Date'].dt.year
# Filter DataFrame for the desired year
df_filtered = df[df['year'] == target_year]
# Check if data for the target year is available
if df_filtered.empty:
st.warning(f"Tidak ada data untuk tahun {target_year}.")
return
# Mapping nilai bulan ke nama bulan
bulan_mapping = {
1: f'Januari {target_year}',
2: f'Februari {target_year}',
3: f'Maret {target_year}',
4: f'April {target_year}',
5: f'Mei {target_year}',
6: f'Juni {target_year}',
7: f'Juli {target_year}',
8: f'Agustus {target_year}',
9: f'September {target_year}',
10: f'Oktober {target_year}',
11: f'November {target_year}',
12: f'Desember {target_year}'
}
# Mengganti nilai dalam kolom 'month' menggunakan mapping
df_filtered['month'] = df_filtered['month'].replace(bulan_mapping)
# Menentukan warna untuk setiap kategori dalam kolom 'score'
warna_label = {
'Negatif': '#FF9AA2',
'Netral': '#FFDAC1',
'Positif': '#B5EAD7'
}
# Sorting unique scores
unique_label = sorted(df_filtered['label'].unique())
# Ensure months are in the correct order
months_order = [
f'Januari {target_year}', f'Februari {target_year}', f'Maret {target_year}', f'April {target_year}', f'Mei {target_year}', f'Juni {target_year}',
f'Juli {target_year}', f'Agustus {target_year}', f'September {target_year}', f'Oktober {target_year}', f'November {target_year}', f'Desember {target_year}'
]
# Sort DataFrame based on the custom order of months
df_filtered['month'] = pd.Categorical(df_filtered['month'], categories=months_order, ordered=True)
df_filtered = df_filtered.sort_values('month')
# Create a bar chart with stacking and manual colors
st.bar_chart(
df_filtered.groupby(['month', 'label']).size().unstack().fillna(0),
color=[warna_label[label] for label in unique_label]
)
@st.cache_data(show_spinner = 'On progress, please wait...')
def all_data_process(texts, df, lookp_dict, stop_words, _sentiment_model, _tfidf_vectorizer):
results = []
analisis = False
if 'Text' in df.columns:
if 'Date' in df.columns:
for text, date in zip(texts, df['Date']):
sentiment_label, emoticon = predict_sentiment(text, _sentiment_model, _tfidf_vectorizer, lookp_dict)
cleaned_text = clean_text(text)
norm_slang_text = normalize_slang(cleaned_text, lookp_dict)
tanpa_stopwords = remove_stopwords(norm_slang_text, stop_words)
result_entry = {
'Date': date,
'Text': text,
'cleaned-text': cleaned_text,
'normalisasi-text': norm_slang_text,
'stopwords-remove': tanpa_stopwords,
'label': sentiment_label,
'emotikon': emoticon,
}
results.append(result_entry)
analisis = True
else:
for text in texts:
sentiment_label, emoticon = predict_sentiment(text, _sentiment_model, _tfidf_vectorizer, lookp_dict)
cleaned_text = clean_text(text)
norm_slang_text = normalize_slang(cleaned_text, lookp_dict)
tanpa_stopwords = remove_stopwords(norm_slang_text, stop_words)
result_entry = {
'Text': text,
'cleaned-text': cleaned_text,
'normalisasi-text': norm_slang_text,
'stopwords-remove': tanpa_stopwords,
'label': sentiment_label,
'emotikon': emoticon,
}
results.append(result_entry)
analisis = True
else:
st.warning("Berkas XLSX harus memiliki kolom bernama 'Text' untuk analisis sentimen.")
return results, analisis
# Fungsi untuk membuat tautan unduhan
def get_table_download_link(df, download_format):
if download_format == "XLSX":
df.to_excel("hasil_sentimen.xlsx", index=False)
return f'Unduh File XLSX'
else:
csv = df.to_csv(index=False)
return f'Unduh File CSV'
# Judul
st.image('https://github.com/naufalnashif/sentiment-analysis-biskita/blob/main/assets/stk_logo-2.jpg?raw=true')
st.title("Sentiment Analysis : Based on Tweets Biskita Transpakuan Bogor 2022-2023")
preference_barchart_date = False
#-----------------------------------------------------General Settings---------------------------------------------------------------
with st.sidebar :
st.subheader('Settings :')
with st.expander("General Settings :"):
# Tambahkan widget untuk memilih model
selected_model = st.selectbox("Pilih Model Sentimen:", ("Ensemble", "Naive Bayes", "Logistic Regression", "Transformer"))
# Pilihan input teks manual atau berkas XLSX
input_option = st.radio("Pilih metode input:", ("Teks Manual", "Unggah Berkas XLSX"))
if input_option == "Teks Manual":
# Input teks dari pengguna
user_input = st.text_area("Masukkan teks:", "")
else:
# Input berkas XLSX
uploaded_file = st.file_uploader("Unggah berkas XLSX", type=["xlsx"])
st.caption("Pastikan berkas XLSX Anda memiliki kolom yang bernama :blue[Text] _(Maks.10000 data)_.")
st.caption("Jika terdapat kolom type :blue[datetime], ganti nama kolom menjadi :blue[Date]")
if uploaded_file is not None:
df = pd.read_excel(uploaded_file)
df = df[:10000]
if 'Text' not in df.columns:
st.warning("Berkas XLSX harus memiliki kolom bernama 'Text' untuk analisis sentimen.")
if not df['Text'].empty:
st.warning("Kolom 'Text' harus mempunyai value.")
else:
texts = df['Text'] # Sesuaikan dengan nama kolom di berkas XLSX Anda
if "Date" in df.columns :
if not df['Date'].empty:
dates = df['Date']
preference_barchart_date = True
#-----------------------------------------------------Preference Settings--------------------------------------------------
with st.expander ("Preference Settings:"):
colormap = st.selectbox("Pilih Warna Wordclouds :", ["Greys", "Purples", "Blues", "Greens", "Oranges", "Reds", "YlOrBr", "YlOrRd", "OrRd", "PuRd", "RdPu", "BuPu", "GnBu", "PuBu", "YlGnBu", "PuBuGn", "BuGn", "YlGn"])
if preference_barchart_date == True:
bar = st.selectbox("Pilih Tampilan Bar Chart :", ("Distribusi Kelas", "Distribusi Kelas Berdasarkan Waktu"), index = 0)
df_target_year = df['Date'].astype(str)
target_year = st.selectbox("Pilih Tahun Bar Chart :", df_target_year.str[:4].unique())
st.info('Tekan "Analysis" kembali jika tampilan menghilang', icon = 'âšī¸')
button = st.button("Analysis")
tab1, tab2, tab3, tab4 = st.tabs(["đ Documentation", "đ Results", "đ¤ĩ Creator", "đ More"])
with tab1:
@st.cache_resource
def tab_1():
st.header("Documentation:")
'''
Langkah - langkah :
1. Buka sidebar sebelah kiri
2. Buka General Settings
3. Pilih Model
4. Pilih Input ('Text Manual', 'File Xlsx')
- Input manual dapat berisi banyak input, lakukan dengan tekan 'enter' untuk menambah line baru
5. File xlsx harus memiliki kolom 'Text'
6. Kolom type datetime "%Y-%m-%d %H:%M:%S" harus bernama 'Date', untuk mengaktifkan fitur tambahan
7. Buka Preferences Settings untuk menyetel tampilan Wordclouds/Barchart
8. Klik Analysis
9. Buka tab Results
'''
st.write('Data bisa dicari di sini:')
more1, more2, more3 = st.columns(3)
with more1 :
st.image('playstore.png', caption = 'Scraping Playstore Reviews')
more1_link = "https://huggingface.co/spaces/naufalnashif/scraping-playstore-reviews"
st.markdown(f"[{more1_link}]({more1_link})")
with more2 :
st.image('News.png', caption = 'Scraping News Headline')
more2_link = "https://huggingface.co/spaces/naufalnashif/scraping-news-headline"
st.markdown(f"[{more2_link}]({more2_link})")
with more3 :
st.image('Ecommerce.png', caption = 'Scraping Ecommerce Product')
more3_link = "https://huggingface.co/spaces/naufalnashif/scraping-ecommerce-2023"
st.markdown(f"[{more3_link}]({more3_link})")
tab_1()
with tab2:
st.header("Results:")
kamus_path = '_json_colloquial-indonesian-lexicon (1).txt'
kamus_sendiri_path = 'kamus_gaul_custom.txt'
tfidf_model_path = 'X_tfidf_model.joblib'
lookp_dict, stop_words, tfidf_vectorizer, model_ensemble, model_nb, model_lr = load_file(kamus_path, kamus_sendiri_path)
sentiment_model = select_sentiment_model(selected_model, model_ensemble, model_lr, model_nb)
# Analisis sentimen
results = []
analisis = False
if input_option == "Teks Manual" and user_input:
if button:
# Pisahkan teks yang dimasukkan pengguna menjadi baris-baris terpisah
user_texts = user_input.split('\n')
for text in user_texts:
sentiment_label, emoticon = predict_sentiment(text, sentiment_model, tfidf_vectorizer, lookp_dict)
cleaned_text = clean_text(text)
norm_slang_text = normalize_slang(cleaned_text, lookp_dict)
tanpa_stopwords = remove_stopwords(norm_slang_text, stop_words)
results.append({
'Text': text,
'cleaned-text' : cleaned_text,
'normalisasi-text' : norm_slang_text,
'stopwords-remove' : tanpa_stopwords,
'label' : sentiment_label,
'emotikon' : emoticon,
})
analisis = True
elif input_option == "Unggah Berkas XLSX" and uploaded_file is not None:
if button:
results, analisis = all_data_process(texts, df, lookp_dict, stop_words, sentiment_model, tfidf_vectorizer)
if results and analisis == True:
df_results = pd.DataFrame(results)
# Membagi tampilan menjadi dua kolom
columns = st.columns(2)
# Kolom pertama untuk Word Cloud
with columns[0]:
st.write("Wordclouds:")
all_texts = [result['stopwords-remove'] for result in results if result['stopwords-remove'] is not None and not pd.isna(result['stopwords-remove'])]
all_texts = " ".join(all_texts)
if all_texts:
wordcloud = WordCloud(width=800, height=660, background_color='white',
colormap=colormap, # Warna huruf
contour_color='black', # Warna kontur
contour_width=2, # Lebar kontur
mask=None, # Gunakan mask untuk bentuk kustom
).generate(all_texts)
st.image(wordcloud.to_array())
else:
st.write("Tidak ada data untuk ditampilkan dalam Word Cloud.")
if 'Date' in df_results.columns:
if bar == "Distribusi Kelas Berdasarkan Waktu":
if not df_results['Date'].empty:
with columns[1]:
buat_chart(df_results, target_year)
else :
# Kolom kedua untuk Bar Chart
with columns[1]:
st.write("Bar Chart :")
# Membuat bar chart
st.bar_chart(
df_results["label"].value_counts()
)
else :
# Kolom kedua untuk Bar Chart
with columns[1]:
st.write("Bar Chart :")
# Membuat bar chart
st.bar_chart(
df_results["label"].value_counts()
)
# Menampilkan hasil analisis sentimen dalam kotak yang dapat diperluas
with st.expander("Hasil Analisis Sentimen"):
# Tampilkan tabel hasil analisis sentimen
st.write(pd.DataFrame(results))
if results:
# Simpan DataFrame ke dalam file CSV
df = pd.DataFrame(results)
csv = df.to_csv(index=False)
# Tampilkan tombol unduh CSV
st.download_button(label="Unduh CSV", data=csv, key="csv_download", file_name="hasil_sentimen.csv")
else:
st.write("Tidak ada data untuk diunduh.")
else:
st.write("Tidak ada data untuk ditampilkan")
with tab3:
@st.cache_resource
def tab_3():
st.header("Profile:")
st.image('https://github.com/naufalnashif/naufalnashif.github.io/blob/main/assets/img/my-profile-semhas.jpeg?raw=true', caption='Naufal Nashif')
st.subheader('Hello, nice to meet you !')
# Tautan ke GitHub
github_link = "https://github.com/naufalnashif/"
st.markdown(f"GitHub: [{github_link}]({github_link})")
# Tautan ke Instagram
instagram_link = "https://www.instagram.com/naufal.nashif/"
st.markdown(f"Instagram: [{instagram_link}]({instagram_link})")
# Tautan ke Website
website_link = "https://naufalnashif.netlify.app/"
st.markdown(f"Website: [{website_link}]({website_link})")
tab_3()
with tab4:
@st.cache_resource
def tab_4():
st.header("More:")
more1, more2, more3 = st.columns(3)
with more1 :
st.image('playstore.png', caption = 'Scraping Playstore Reviews')
more1_link = "https://huggingface.co/spaces/naufalnashif/scraping-playstore-reviews"
st.markdown(f"[{more1_link}]({more1_link})")
with more2 :
st.image('News.png', caption = 'Scraping News Headline')
more2_link = "https://huggingface.co/spaces/naufalnashif/scraping-news-headline"
st.markdown(f"[{more2_link}]({more2_link})")
with more3 :
st.image('Ecommerce.png', caption = 'Scraping Ecommerce Product')
more3_link = "https://huggingface.co/spaces/naufalnashif/scraping-ecommerce-2023"
st.markdown(f"[{more3_link}]({more3_link})")
tab_4()
# Garis pemisah
st.divider()
st.write('Thank you for trying the demo!')
st.caption('Best regards, Naufal Nashif :sunglasses: | Šī¸ 2023')