A movie recommendation web-app
You can test this app on -> Streamlit
We take text samples as an input(in this case, it's the plots of the movies) and extract N number of keywords from each of them.(I used 25 in this notebook, because of the kaggle's limitations. I observed, the more is the better). This N keywords represents our whole plot. Then we convert these keywords into sentence embeddings using Sentence Tokenizer(There are bunch of methods which can turn words into embeddings, but I observed Sentence Tokenizer worked great).
For this demo I was need a dataset, which contain both movie plots and movie titles. I found this dataset, which contains decent amount of unique movies(and also their plot).