Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import transformers
|
3 |
+
import torch
|
4 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
5 |
+
|
6 |
+
# Define the paths of the pre-trained models
|
7 |
+
model1_path = "saisi/finetuned-Sentiment-classfication-ROBERTA-Base-model"
|
8 |
+
model2_path = "saisi/finetuned-Sentiment-classfication-DISTILBERT-model"
|
9 |
+
|
10 |
+
# Initialize the tokenizer and models for sentiment analysis
|
11 |
+
tokenizer1 = AutoTokenizer.from_pretrained(model1_path)
|
12 |
+
model1 = AutoModelForSequenceClassification.from_pretrained(model1_path)
|
13 |
+
tokenizer2 = AutoTokenizer.from_pretrained(model2_path)
|
14 |
+
model2 = AutoModelForSequenceClassification.from_pretrained(model2_path)
|
15 |
+
|
16 |
+
# Define a function to preprocess the text data
|
17 |
+
def preprocess(text):
|
18 |
+
new_text = []
|
19 |
+
# Replace user mentions with '@user'
|
20 |
+
for t in text.split(" "):
|
21 |
+
t = '@user' if t.startswith('@') and len(t) > 1 else t
|
22 |
+
# Replace links with 'http'
|
23 |
+
t = 'http' if t.startswith('http') else t
|
24 |
+
new_text.append(t)
|
25 |
+
# Join the preprocessed text
|
26 |
+
return " ".join(new_text)
|
27 |
+
|
28 |
+
# Define a function to perform sentiment analysis on the input text using model 1
|
29 |
+
def sentiment_analysis_model1(text):
|
30 |
+
# Preprocess the input text
|
31 |
+
text = preprocess(text)
|
32 |
+
# Tokenize the input text using the pre-trained tokenizer
|
33 |
+
encoded_input = tokenizer1(text, return_tensors='pt')
|
34 |
+
# Feed the tokenized input to the pre-trained model and obtain output
|
35 |
+
output = model1(**encoded_input)
|
36 |
+
# Obtain the prediction scores for the output
|
37 |
+
scores_ = output[0][0].detach().numpy()
|
38 |
+
# Apply softmax activation function to obtain probability distribution over the labels
|
39 |
+
scores_ = torch.nn.functional.softmax(torch.from_numpy(scores_), dim=0).numpy()
|
40 |
+
# Format the output dictionary with the predicted scores
|
41 |
+
labels = ['Negative', 'Positive']
|
42 |
+
scores = {l:float(s) for (l,s) in zip(labels, scores_) }
|
43 |
+
# Return the scores
|
44 |
+
return scores
|
45 |
+
|
46 |
+
# Define a function to perform sentiment analysis on the input text using model 2
|
47 |
+
def sentiment_analysis_model2(text):
|
48 |
+
# Preprocess the input text
|
49 |
+
text = preprocess(text)
|
50 |
+
# Tokenize the input text using the pre-trained tokenizer
|
51 |
+
encoded_input = tokenizer2(text, return_tensors='pt')
|
52 |
+
# Feed the tokenized input to the pre-trained model and obtain output
|
53 |
+
output = model2(**encoded_input)
|
54 |
+
# Obtain the prediction scores for the output
|
55 |
+
scores_ = output[0][0].detach().numpy()
|
56 |
+
# Apply softmax activation function to obtain probability distribution over the labels
|
57 |
+
scores_ = torch.nn.functional.softmax(torch.from_numpy(scores_), dim=0).numpy()
|
58 |
+
# Format the output dictionary with the predicted scores
|
59 |
+
labels = ['Negative', 'Neutral', 'Positive']
|
60 |
+
scores = {l:float(s) for (l,s) in zip(labels, scores_) }
|
61 |
+
# Return the scores
|
62 |
+
return scores
|
63 |
+
|
64 |
+
# Define the Streamlit app
|
65 |
+
def app():
|
66 |
+
# Define the app title
|
67 |
+
st.title("Sentiment Analysis")
|
68 |
+
# Define the input field
|
69 |
+
text_input = st.text_input("Enter text:")
|
70 |
+
# Define the model selection dropdown
|
71 |
+
model_selection = st.selectbox("Select a model:", ["Model 1", "Model 2"])
|
72 |
+
# Perform sentiment analysis when the submit button is clicked
|
73 |
+
if st.button("Submit"):
|
74 |
+
if text_input
|