🎬 CineScore

Home	Suggestions
Review1	Review 2

📝 Project Overview

This project (assignment) is a Flask-based web application that combines LSTM-based sentiment analysis and IMDb web scraping to provide a platform where users can:

1. Search for movies on IMDb database on a text input (e.g., movie name or keyword).
2. View a list of movie options with images fetched from their official IMDb pages.
3. Select a movie and view its user reviews along with ratings.
4. See predicted sentiment scores based on a trained LSTM model for each review.

The app scrapes official IMDb user reviews and computes sentiment scores for each review using a pre-trained LSTM model. It then displays the reviews, user scores, and predicted scores in a table.

⭐ Features

• Movie Search: Users can input a movie or keyword, and the app will search IMDb for relevant movies.
• Image Scraping: The app scrapes movie images from the official IMDb webpage.
• Review Scraping: User reviews are scraped from IMDb's official review page.
• Sentiment Prediction: The app uses a pre-trained LSTM model to predict sentiment scores for each review.
• Table of Results: Display of movie reviews, IMDb user scores and predicted sentiment scores in an organized table.

💻 Installation

1. Clone the Repository:

git clone https://github.com/your-username/movie-sentiment-analysis-flask.git
cd movie-sentiment-analysis-flask

2. Create a Virtual Environment (optional but recommended):

python3 -m venv venv
source venv/bin/activate

3. Installation:

pip install -r requirements.txt

🚀 Running the App

To start the Flask server, run the following command:

python app.py

📚 Datasets and Resources

1. IMDB Dataset for Sentiment Analysis

• Description: The IMDB dataset contains movie reviews and sentiment labels (positive/negative). It's used to train the LSTM model for predicting sentiment scores for movie reviews.
• Source: IMDB Dataset

2. GloVe Embeddings for Word Representation

• Description: GloVe (Global Vectors for Word Representation) is used to convert words into word vectors. The 100-dimensional word vectors are pre-trained on a large corpus of text and are used to initialize the embedding layer in the LSTM model.
• Source: GloVe 6B 100-dimensional Embeddings

3. IMDb Website for Scraping Movie Data and Reviews

• Description: IMDb’s official website is used for scraping movie information, including movie names, images, and user reviews. This data is then presented to users in the app.
• URL: IMDb (duh!)

4. IMDb Official Database for Movie Names

• Description: The official IMDb dataset is used to search for movie names and tokens. It provides detailed information about movies, including their names, genres, and other metadata. The data is used to fetch matching movie names based on user input.
• Source: IMDb Official Datasets

📁 Project Structure

CineScore
├── app.py
├── download_nltk_resources.py
│
├── helpers
│   └── helper.py
│
├── static
│   ├── css
│   │   └── style.css
│   ├── db
│   │   ├── movie.csv
|   |   └── lsmt123.keras 
│   └── images
│       └── no.svg
│
└── templates
    ├── index.html
    ├── noreview.html
    ├── review.html
    └── suggestions.html

⚙️ How It Works

1. First, the user provides their favorite movie as input.

Example: Interstellar

2. The application then fetches all the rows that match the movie name from the IMDb database.

For example:

| tconst    | primaryTitle |
|-----------|--------------|
| tt0816692 | Interstellar |

3. Here, tconst is important because it helps to scrape the specific webpage for images and reviews.

The basic IMDb URL is: 
https://www.imdb.com/title/{tconst}

So, for the above example, the URL will be:
https://www.imdb.com/title/tt0816692

4. After retrieving all the URL's, the program scrapes the movie poster by going to each page. It uses BeautifulSoup4 to search for the image div using the get_img function (under helper.py).

5. The application then shows the user all the movie results that match their input. The user picks their favorite movie among the queries.

6. Afterward, it scrapes all the user reviews and their scores for the selected movie using the get_reviews function.

The URL format for reviews is:
https://www.imdb.com/title/{tconst}/reviews/

For example:
https://www.imdb.com/title/tt0816692/reviews

7. After collecting the reviews, each review is passed through a preprocessing function (the same function used for training the model). Then, the predict function is used to generate the predicted scores, which are displayed in a table.

| Review                 | User Score | Predicted Score |
|------------------------|------------|-----------------|
| This movie is the best |    10      |       10        |

💡 Room for Improvements

1. Use Database

• You could use database using SQLAlchemy for getting suggestions in search bar as user type.

2. Better Search Functionality

• Maybe add additional parameters, such as genre, release year, or IMDb rating.

3. Apply your own Ideas

• Bring your unique vision/ideas and enhancements into copying and adding into this project.

🚨 Common Errors & Solutions

• If you get error tensorflow not found while installing tensorflow or something related to it check python version and its compatibility

  • As of right now the latest version of python is 3.13 but the latest version of tensorflow==2.10 only supports till python==3.10
  • You can manage python version using
    • pyenv for linux
    • conda for windows

• Also if you want to use CUDA check its version matches with python and tensorflow supported compatibility

  You can check compatible version here: https://www.tensorflow.org/install/source_windows#gpu