Create app.py

app.py

@@ -0,0 +1,188 @@
+import streamlit as st
+
+import transformers
+
+import torch
+
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+
+
+
+# Define the paths of the pre-trained models
+
+model1_path = "Winnie-Kay/Finetuned_BertModel_SentimentAnalysis"
+
+model2_path = "Winnie-Kay/Sentiment-Analysis-Roberta-bases"
+
+
+
+
+# Initialize the tokenizer and models for sentiment analysis
+
+tokenizer1 = AutoTokenizer.from_pretrained(model1_path)
+
+model1 = AutoModelForSequenceClassification.from_pretrained(model1_path)
+
+tokenizer2 = AutoTokenizer.from_pretrained(model2_path)
+
+model2 = AutoModelForSequenceClassification.from_pretrained(model2_path)
+
+
+
+
+# Define a function to preprocess the text data
+
+def preprocess(text):
+
+    new_text = []
+
+    # Replace user mentions with '@user'
+
+    for t in text.split(" "):
+
+        t = '@user' if t.startswith('@') and len(t) > 1 else t
+
+        # Replace links with 'http'
+
+        t = 'http' if t.startswith('http') else t
+
+        new_text.append(t)
+
+    # Join the preprocessed text
+
+    return " ".join(new_text)
+
+
+
+
+# Define a function to perform sentiment analysis on the input text using model 1
+
+def sentiment_analysis_model1(text):
+
+    # Preprocess the input text
+
+    text = preprocess(text)
+
+
+
+
+    # Tokenize the input text using the pre-trained tokenizer
+
+    encoded_input = tokenizer1(text, return_tensors='pt')
+
+   
+
+    # Feed the tokenized input to the pre-trained model and obtain output
+
+    output = model1(**encoded_input)
+
+   
+
+    # Obtain the prediction scores for the output
+
+    scores_ = output[0][0].detach().numpy()
+
+   
+
+    # Apply softmax activation function to obtain probability distribution over the labels
+
+    scores_ = torch.nn.functional.softmax(torch.from_numpy(scores_), dim=0).numpy()
+
+   
+
+    # Format the output dictionary with the predicted scores
+
+    labels = ['Negative', 'Positive']
+
+    scores = {l:float(s) for (l,s) in zip(labels, scores_) }
+
+   
+
+    # Return the scores
+
+    return scores
+
+
+
+
+# Define a function to perform sentiment analysis on the input text using model 2
+
+def sentiment_analysis_model2(text):
+
+    # Preprocess the input text
+
+    text = preprocess(text)
+
+
+
+
+    # Tokenize the input text using the pre-trained tokenizer
+
+    encoded_input = tokenizer2(text, return_tensors='pt')
+
+   
+
+    # Feed the tokenized input to the pre-trained model and obtain output
+
+    output = model2(**encoded_input)
+
+   
+
+    # Obtain the prediction scores for the output
+
+    scores_ = output[0][0].detach().numpy()
+
+   
+
+    # Apply softmax activation function to obtain probability distribution over the labels
+
+    scores_ = torch.nn.functional.softmax(torch.from_numpy(scores_), dim=0).numpy()
+
+   
+
+    # Format the output dictionary with the predicted scores
+
+    labels = ['Negative', 'Neutral', 'Positive']
+
+    scores = {l:float(s) for (l,s) in zip(labels, scores_) }
+
+   
+
+    # Return the scores
+
+    return scores
+
+
+
+
+# Define the Streamlit app
+
+def app():
+
+    # Define the app title
+
+    st.title("Sentiment Analysis")
+
+
+
+
+    # Define the input field
+
+    text_input = st.text_input("Enter text:")
+
+
+
+
+    # Define the model selection dropdown
+
+    model_selection = st.selectbox("Select a model:", ["Model 1", "Model 2"])
+
+
+
+
+    # Perform sentiment analysis when the submit button is clicked
+
+    if st.button("Submit"):
+
+        if text_input