Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import pipeline
|
3 |
+
import pandas as pd
|
4 |
+
|
5 |
+
# Pesan saat startup untuk memastikan pipeline mulai dimuat.
|
6 |
+
print("Memuat model pipeline...")
|
7 |
+
|
8 |
+
# -----------------------------------------------------------------------------
|
9 |
+
# 1. Muat semua model pipeline saat aplikasi dimulai.
|
10 |
+
# Ini adalah praktik terbaik agar model siap saat input pertama masuk.
|
11 |
+
# -----------------------------------------------------------------------------
|
12 |
+
try:
|
13 |
+
# Model 1: DistilBERT
|
14 |
+
pipe_distilbert = pipeline("text-classification", model="distilbert-base-uncased-finetuned-sst-2-english")
|
15 |
+
|
16 |
+
# Model 2: BERT
|
17 |
+
pipe_bert = pipeline("text-classification", model="gchhablani/bert-base-cased-finetuned-sst2")
|
18 |
+
|
19 |
+
# Model 3: RoBERTa
|
20 |
+
pipe_roberta = pipeline("text-classification", model="cardiffnlp/twitter-roberta-base-sentiment")
|
21 |
+
|
22 |
+
print("Semua model berhasil dimuat.")
|
23 |
+
except Exception as e:
|
24 |
+
print(f"Error saat memuat model: {e}")
|
25 |
+
# Jika model gagal dimuat, aplikasi tidak akan bisa berjalan.
|
26 |
+
# Anda bisa menambahkan penanganan error yang lebih baik di sini.
|
27 |
+
|
28 |
+
# Mapping label untuk model RoBERTa agar lebih mudah dibaca.
|
29 |
+
roberta_label_map = {
|
30 |
+
"LABEL_0": "NEGATIVE",
|
31 |
+
"LABEL_1": "NEUTRAL",
|
32 |
+
"LABEL_2": "POSITIVE"
|
33 |
+
}
|
34 |
+
|
35 |
+
# -----------------------------------------------------------------------------
|
36 |
+
# 2. Buat fungsi utama untuk prediksi.
|
37 |
+
# Fungsi ini akan menerima input teks dan mengembalikan hasil dari ketiga model.
|
38 |
+
# -----------------------------------------------------------------------------
|
39 |
+
def predict_sentiments(text):
|
40 |
+
"""
|
41 |
+
Menerima input teks, melakukan prediksi dengan tiga model,
|
42 |
+
dan mengembalikan hasilnya dalam format yang bisa ditampilkan oleh Gradio.
|
43 |
+
"""
|
44 |
+
if not text.strip():
|
45 |
+
# Jika tidak ada teks, kembalikan dataframe kosong
|
46 |
+
return pd.DataFrame(columns=["Model", "Label", "Confidence Score"])
|
47 |
+
|
48 |
+
results = []
|
49 |
+
|
50 |
+
# Prediksi dengan DistilBERT
|
51 |
+
pred_distilbert = pipe_distilbert(text)[0]
|
52 |
+
results.append(["DistilBERT", pred_distilbert['label'], f"{pred_distilbert['score']:.4f}"])
|
53 |
+
|
54 |
+
# Prediksi dengan BERT
|
55 |
+
pred_bert = pipe_bert(text)[0]
|
56 |
+
results.append(["BERT", pred_bert['label'], f"{pred_bert['score']:.4f}"])
|
57 |
+
|
58 |
+
# Prediksi dengan RoBERTa
|
59 |
+
pred_roberta = pipe_roberta(text)[0]
|
60 |
+
# Ganti label RoBERTa sesuai mapping
|
61 |
+
label_roberta = roberta_label_map.get(pred_roberta['label'], pred_roberta['label'])
|
62 |
+
results.append(["RoBERTa", label_roberta, f"{pred_roberta['score']:.4f}"])
|
63 |
+
|
64 |
+
# Buat DataFrame dari hasil untuk output tabel yang rapi
|
65 |
+
df = pd.DataFrame(results, columns=["Model", "Label", "Confidence Score"])
|
66 |
+
return df
|
67 |
+
|
68 |
+
# -----------------------------------------------------------------------------
|
69 |
+
# 3. Buat dan jalankan antarmuka Gradio.
|
70 |
+
# -----------------------------------------------------------------------------
|
71 |
+
demo = gr.Interface(
|
72 |
+
fn=predict_sentiments,
|
73 |
+
inputs=gr.Textbox(
|
74 |
+
lines=5,
|
75 |
+
placeholder="Masukkan teks di sini untuk dianalisis sentimennya...",
|
76 |
+
label="Input Teks"
|
77 |
+
),
|
78 |
+
outputs=gr.Dataframe(
|
79 |
+
headers=["Model", "Label", "Confidence Score"],
|
80 |
+
datatype=["str", "str", "str"],
|
81 |
+
label="Hasil Prediksi"
|
82 |
+
),
|
83 |
+
title="🤖 Perbandingan Model Analisis Sentimen",
|
84 |
+
description="Aplikasi ini membandingkan hasil prediksi sentimen dari tiga model transformer yang berbeda: DistilBERT, BERT, dan RoBERTa. Cukup masukkan teks dan klik 'Submit'.",
|
85 |
+
allow_flagging="never",
|
86 |
+
examples=[
|
87 |
+
["I love the new features, this is an amazing update!"],
|
88 |
+
["The performance is a bit slow and sometimes it crashes."],
|
89 |
+
["This product is okay, not great but not terrible either."]
|
90 |
+
]
|
91 |
+
)
|
92 |
+
|
93 |
+
# Jalankan aplikasi
|
94 |
+
if __name__ == "__main__":
|
95 |
+
demo.launch()
|