The goal of the project is to predict and label a movie plot summary with its appropriate genre label. project is based on Recurrent Neural Networks using LSTM units for labelling.
Description -
The project is web application made using Flask web framework to provide an interface to access the trained R-NN model, the app takes in a plot as the input parameter and predicts the genre of the movie plot. The application on its initial run would take sometime time to train the network locally and stores the trained model within its local file system to be accessed to for later prediction, where we would used the already trained model to predict the genre of the plot provided.
we have used Glove Embeddings to encode the words into a vector space to be trained using the R-NN model. Refer Glove embeddings to better understand and explore.
DateSet -
The dataset can be downloaded from - ftp://ftp.fu-berlin.de/pub/misc/movies/database/frozendata/ , one would need to parse the file inorder to properly attribute dataset with titles, plot and genres. The parsing activity is already done and the respective csv file attached in the repository.
made use of Flask to provide a basic web application interface as a wrapper for the RNN model.
Libraries used -
Keras 2.0
Flask - Micro webframework for python
Results -
The model has been able to achieve an accuracy of 74% on validation dataset.
Acheived an F1 score of 0.91 on validation set
Usage -
Run the program using - $ python start.py
Use endpoint: http://localhost/node/7777