Upload 3 files

.streamlit/config.toml

@@ -0,0 +1,3 @@
+[server]
+runOnSave = true
+fileWatcherType = "poll"

streamlit_app.py/Homepage.py

@@ -1,22 +1,46 @@
 import streamlit as st
+from st_pages import Page, show_pages
 
-st.set_page_config(
-page_title="Sentiment Detection"
+st.set_page_config(page_title="Sentiment Detection", page_icon="🏠")
+
+show_pages(
+    [
+        Page("streamlit_app.py/Homepage.py", "Home", "🏠"),
+        Page(
+            "streamlit_app.py/pages/Sentiment_Detection.py", "Sentiment Detection", "📝"
+        ),
+    ]
+)
+
+st.title("Final Project in Machine Learning Course - Sentiment Detection")
+st.markdown(
+    """
+    **Team members:**
+    | Student ID | Full Name                |
+    | ---------- | ------------------------ |
+    | 19120600   | Bùi Nguyên Nghĩa         |
+    | 20120089   | Lê Xuân Hoàng            |
+    | 20120422   | Nguyễn Thị Ánh Tuyết     |
+    | 20120460   | Lê Nguyễn Hải Dương      |
+    | 20120494   | Lê Xuân Huy              |
+    """
 )
-   
-st.title("Sentiment Detection")
-    
-st.sidebar.success("Select a page above.")
 
 st.header("The Need for Sentiment Detection")
-st.text("""
-        Spam detection algorithms are used to detect and filter junk and spam emails with a high level of accuracy.
-        It is said that around half of all emails are spam, depending on the user. These emails can include scams or viruses intended to cause harm.
-        """)
+st.markdown(
+    """
+    Sentiment detection algorithms are used to detect sentiment in a comment or a review.
+    It is said that around 90% of consumers read online reviews before visiting a business or buying a product.
+    These reviews can be positive or negative or neutral, and it is important to know what the customers are saying about your business.
+    """
+)
+
+st.header("Technology used")
+st.markdown(
+    """
+    In this demo, we used BERT as the model for sentiment detection. BERT is a transformer-based model that was proposed in 2018 by Google.
+    It is a pre-trained model that can be used for various NLP tasks such as sentiment detection, question answering, etc.
+    """
+)
+
 
-st.header("Data Source")
-st.text("""
-        Data Source: Preprocessed TREC 2007 Public Corpus Dataset. 
-        Link: https://www.kaggle.com/datasets/imdeepmind/preprocessed-trec-2007-public-corpus-dataset
-        
-        """)

streamlit_app.py/pages/Sentiment_Detection.py

@@ -0,0 +1,117 @@
+from os import path
+import streamlit as st
+
+# import pickle
+
+# from tensorflow import keras
+import tensorflow as tf
+import torch
+from torch import nn
+from transformers import BertModel, BertTokenizer
+
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+MODEL_NAME = "bert-base-cased"
+MODEL_PATH = path.join(path.dirname(__file__), "bert_model.h5")
+
+
+# Build the Sentiment Classifier class
+class SentimentClassifier(nn.Module):
+    # Constructor class
+    def __init__(self, n_classes):
+        super(SentimentClassifier, self).__init__()
+        self.bert = BertModel.from_pretrained(MODEL_NAME)
+        self.drop = nn.Dropout(p=0.3)
+        self.out = nn.Linear(self.bert.config.hidden_size, n_classes)
+
+    # Forward propagaion class
+    def forward(self, input_ids, attention_mask):
+        _, pooled_output = self.bert(
+            input_ids=input_ids, attention_mask=attention_mask, return_dict=False
+        )
+        #  Add a dropout layer
+        output = self.drop(pooled_output)
+        return self.out(output)
+
+
+@st.cache_resource
+def load_model_and_tokenizer():
+    model = SentimentClassifier(3)
+    model.load_state_dict(torch.load(MODEL_PATH, map_location=torch.device("cpu")))
+    model.eval()
+    return model, BertTokenizer.from_pretrained("bert-base-cased")
+
+
+def predict(content):
+    model, tokenizer = load_model_and_tokenizer()
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+    encoded_review = tokenizer.encode_plus(
+        content,
+        max_length=160,
+        add_special_tokens=True,
+        return_token_type_ids=False,
+        pad_to_max_length=True,
+        return_attention_mask=True,
+        return_tensors="pt",
+    )
+
+    input_ids = encoded_review["input_ids"].to(device)
+    attention_mask = encoded_review["attention_mask"].to(device)
+
+    output = model(input_ids, attention_mask)
+    _, prediction = torch.max(output, dim=1)
+
+    class_names = ["negative", "neutral", "positive"]
+
+    return class_names[prediction]
+
+
+def main():
+    st.set_page_config(page_title="Sentiment Detection", page_icon="📝")
+
+    # giving a title to our page
+    st.title("Sentiment detection")
+    contents = st.text_area(
+        "Please enter reviews/sentiment/setences/contents:",
+        placeholder="Enter your text here",
+        height=200,
+    )
+
+    prediction = ""
+
+    # Create a prediction button
+    if st.button("Analyze Sentiment"):
+        stripped = contents.strip()
+        if not stripped:
+            st.error("Please enter some text.")
+            return
+
+        prediction = predict(contents)
+        if prediction == "positive":
+            st.success("This is positive 😄")
+        elif prediction == "negative":
+            st.error("This is negative 😟")
+        else:
+            st.warning("This is neutral 🙂")
+
+    upload_file = st.file_uploader("Or upload a file", type=["txt"])
+    if upload_file is not None:
+        contents = upload_file.read().decode("utf-8")
+
+        for line in contents.splitlines():
+            line = line.strip()
+            if not line:
+                continue
+
+            prediction = predict(line)
+            if prediction == "positive":
+                st.success(line + "\n\nThis is positive 😄")
+            elif prediction == "negative":
+                st.error(line + "\n\nThis is negative 😟")
+            else:
+                st.warning(line + "\n\nThis is neutral 🙂")
+
+
+if __name__ == "__main__":
+    main()