Spaces:
Runtime error
Runtime error
import streamlit as st | |
import plotly as plt | |
import numpy as np | |
import pandas as pd | |
import webbrowser | |
import requests | |
import json | |
from streamlit_lottie import st_lottie | |
st.set_page_config(page_title= "Welcome Page", page_icon ="π") | |
st.sidebar.success("Select The Page You Want to Explore: ") | |
st.title("Welcome to my Sentiment Analysis App") | |
def load_lottiefile(filepath: str): | |
with open(filepath, "r") as f: | |
return json.load(f) | |
# initializaing my session state | |
if 'lottie_hello' not in st.session_state: | |
st.session_state.lottie_hello = load_lottiefile("./lottie_animations/main.json") | |
# creating a funciton to upload the file while implementing session state | |
def handle_uploaded_file(uploaded_file): | |
if uploaded_file is not None: | |
st.session_state.lottie_hello = load_lottiefile(uploaded_file.name) | |
# displaying the Lottie animation | |
st_lottie(st.session_state.lottie_hello, height=200) | |
st.markdown("""On this app, you will be able to classify Movie Review sentiments with the Tiny-Bert model | |
The objective of this challenge is to develop a machine learning model to assess if a twitter post that is related to vaccinations is positive or negative.""") | |
st.subheader("""Variable Definition:""") | |
st.write(""" | |
**Review File**: Unique identifier of the review | |
**Content**: Text contained in the review the user gave | |
**Sentiment**: Sentiment of the review (Positive and Negative, Or 0 for Negative, 1 for positive) | |
**Train.csv**: Labelled tweets on which to train your model | |
The Models I fine-tuned include: \n | |
- Roberta: Achieving an Accuracy score of 0.94 but did overfit \n | |
- Tiny Bert: Achieving an Accuracy scrore of 0.87 barely overfitted | |
""") | |
data= pd.read_csv("datasets/Train.csv") | |
st.subheader("A sample of the orginal Dataframe (Train.csv)") | |
st.write(data.head()) | |
st.subheader("A sample of the preprocessed dataset") | |
data_clean= pd.read_csv("datasets/capstone_data.csv") | |
data_clean= data_clean.drop("Unnamed: 0", axis= 1) | |
st.write(data_clean.head()) |