Spaces:
Runtime error
Runtime error
Commit
•
3daec5b
1
Parent(s):
8569b65
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import pandas as pd
|
3 |
+
import numpy as np
|
4 |
+
import joblib
|
5 |
+
import spacy
|
6 |
+
from wordcloud import WordCloud
|
7 |
+
from io import StringIO, BytesIO
|
8 |
+
import mimetypes
|
9 |
+
from transformers import CamembertForSequenceClassification, CamembertTokenizer
|
10 |
+
import torch
|
11 |
+
|
12 |
+
# Model Loading
|
13 |
+
model = joblib.load('model.pkl')
|
14 |
+
vectorizer = joblib.load('vectorizer.pkl')
|
15 |
+
|
16 |
+
camembert_model = CamembertForSequenceClassification.from_pretrained("camembert-base", num_labels=2)
|
17 |
+
state_dict = torch.load('camembertperso.pth', map_location='cpu')
|
18 |
+
camembert_model.load_state_dict(state_dict, strict=False)
|
19 |
+
tokenizer = CamembertTokenizer.from_pretrained('camembert-base', do_lower_case=True)
|
20 |
+
|
21 |
+
nlp = spacy.load("fr_core_news_sm")
|
22 |
+
|
23 |
+
# Text Processing Functions
|
24 |
+
def clean_text(text):
|
25 |
+
return text.strip().lower()
|
26 |
+
|
27 |
+
def lemmatize_text(text):
|
28 |
+
doc = nlp(text)
|
29 |
+
lemmatized_text = " ".join([token.lemma_ for token in doc])
|
30 |
+
return lemmatized_text
|
31 |
+
|
32 |
+
# Prediction Functions
|
33 |
+
def predict_label(text):
|
34 |
+
cleaned_text = clean_text(text)
|
35 |
+
lemmatized_text = lemmatize_text(cleaned_text)
|
36 |
+
|
37 |
+
vectorized_text = vectorizer.transform([lemmatized_text])
|
38 |
+
label = model.predict(vectorized_text)[0]
|
39 |
+
|
40 |
+
probability_score = model.decision_function(vectorized_text)[0]
|
41 |
+
probability = 1 / (1 + np.exp(-probability_score))
|
42 |
+
|
43 |
+
return label, probability
|
44 |
+
|
45 |
+
def predict_camembert(text):
|
46 |
+
tokens = tokenizer.encode_plus(text, return_tensors="pt")
|
47 |
+
|
48 |
+
with torch.no_grad():
|
49 |
+
outputs = camembert_model(**tokens)
|
50 |
+
|
51 |
+
if len(outputs) == 1:
|
52 |
+
logits = outputs[0]
|
53 |
+
else:
|
54 |
+
logits = outputs[1]
|
55 |
+
|
56 |
+
predictions = torch.argmax(logits, dim=1).item()
|
57 |
+
probabilities = torch.softmax(logits, dim=1)[:, 1].item()
|
58 |
+
|
59 |
+
return predictions, probabilities
|
60 |
+
|
61 |
+
# App Interface
|
62 |
+
st.title('Analyse de sentiments')
|
63 |
+
|
64 |
+
st.write('Cet outil permet de prédire si une review est positive ou négative.')
|
65 |
+
|
66 |
+
review_text = st.text_area('Saisir la review ou charger un fichier :')
|
67 |
+
|
68 |
+
if st.button('Prédire et générer le nuage de mots'):
|
69 |
+
# LinearSVC Prediction
|
70 |
+
label_linear_svc, probability_linear_svc = predict_label(review_text)
|
71 |
+
|
72 |
+
# Display LinearSVC Results
|
73 |
+
st.write('Résultats de LinearSVC:')
|
74 |
+
if label_linear_svc == 0:
|
75 |
+
st.write('La review est négative.')
|
76 |
+
else:
|
77 |
+
st.write('La review est positive.')
|
78 |
+
|
79 |
+
# Display LinearSVC Prediction Score
|
80 |
+
st.write('Score de prédiction (LinearSVC) :', f'**{label_linear_svc}**', unsafe_allow_html=True)
|
81 |
+
|
82 |
+
# Display LinearSVC Probability
|
83 |
+
st.write('Probabilité (LinearSVC) :', f'**{probability_linear_svc:.2%}**', unsafe_allow_html=True)
|
84 |
+
|
85 |
+
# CamemBERT Prediction
|
86 |
+
label_camembert, probability_camembert = predict_camembert(review_text)
|
87 |
+
|
88 |
+
# Display CamemBERT Results
|
89 |
+
st.write('Résultats de Camembert:')
|
90 |
+
if label_camembert == 0:
|
91 |
+
st.write('La review est négative.')
|
92 |
+
else:
|
93 |
+
st.write('La review est positive.')
|
94 |
+
|
95 |
+
# Display CamemBERT Prediction Score
|
96 |
+
st.write('Score de prédiction (Camembert) :', f'**{label_camembert}**', unsafe_allow_html=True)
|
97 |
+
|
98 |
+
# Display CamemBERT Probability
|
99 |
+
st.write('Probabilité (Camembert) :', f'**{probability_camembert:.2%}**', unsafe_allow_html=True)
|
100 |
+
|
101 |
+
# Lemmatize and Exclude Stop Words
|
102 |
+
doc = nlp(review_text)
|
103 |
+
lemmatized_text_no_stopwords = " ".join([token.lemma_ for token in doc if not token.is_stop])
|
104 |
+
|
105 |
+
|
106 |
+
# Générer le nuage de mots
|
107 |
+
wordcloud = WordCloud(width=800, height=400, background_color='white').generate(lemmatized_text_no_stopwords)
|
108 |
+
st.image(wordcloud.to_image())
|
109 |
+
|
110 |
+
# Créer un bouton pour l'upload d'un fichier
|
111 |
+
uploaded_file = st.file_uploader("Charger un fichier texte", type=["txt", "csv"])
|
112 |
+
if uploaded_file is not None:
|
113 |
+
content_type, _ = mimetypes.guess_type(uploaded_file.name)
|
114 |
+
if content_type == 'text/plain':
|
115 |
+
file_contents = uploaded_file.read().decode("utf-8")
|
116 |
+
st.text(file_contents)
|
117 |
+
|
118 |
+
# Lemmatiser le texte et exclure les mots vides
|
119 |
+
doc = nlp(file_contents)
|
120 |
+
lemmatized_text_no_stopwords = " ".join([token.lemma_ for token in doc if not token.is_stop])
|
121 |
+
|
122 |
+
# Générer le nuage de mots à partir du fichier uploadé
|
123 |
+
wordcloud = WordCloud(width=800, height=400, background_color='white').generate(lemmatized_text_no_stopwords)
|
124 |
+
st.image(wordcloud.to_image())
|
125 |
+
elif content_type == 'text/csv':
|
126 |
+
df = pd.read_csv(uploaded_file)
|
127 |
+
st.write(df)
|