After reading a book, many readers would like to watch a movie that is very similar to the books they saw online. Book-Movie-Recommender recommends a book based on the input as the summary of the book.
There are two datasets involved:
CMU's Movie Summary Corpus and
CMU's Book Summary Dataset
A summary is chosen from the book dataset, and then it's cosine similarity with a movie summary is calculated.
The cosine similarities are ranked and the 5 highest values are shown.
Brute force, i.e. value = cosine_similarity(Book_Summary, Movie_Summary)
for all Movie_Summary
's takes too much time.
An alternative is to sort all movies by genre's and process it according to genres. However, Anime
is a genre of a
movie that does not exist for book. The last attempt which seems to give approximately good results is to create a relative
cosine_similarity Hash Table. Calculate a the cosine_similarity with respect to the first summary and store the
cos_sim*1000
in a Hash Table. Then check the value*1000
in the hash table.
The current files do not include the Hash Table.
The current files are
Calculate-RelativeCos-Movies.py
which is the relative cosine calculator.Driver-RelativeCos.py
which is the driver to find out the best match.