0
$\begingroup$

I am building a content-based recommendation system for hotel accommodation. I have a hotel name, hotel description and location. I combined hotel name, description and location. Then, applied NLP and converted into vectors using TF-IDF.

I am taking hotel name and location as an input and recommend the hotel based on cosine similarity. I wanted to use word2vec instead of TF-IDF and make a recommendation. Anyone can suggest how to do this?

$\endgroup$
0
$\begingroup$

You can find a number of blogs teaching the use of Word2Vec,

If you are building a hotel recommendation system, I think the name of the hotel is not a good feature. Users are less concerned with the name of the hotel than its location or description provided.

  • You can take a small description from the user and the location in which they would like to search for the hotel. Using Doc2Vec, convert the description ( text ) into a fixed-sized vector ( embedding ). Find a hotel whose description is the most similar to the one provided by the user ( maybe using Cosine Similarity or any other measures ).

  • Before this find a hotel that is close to the location provided by the user ( say a hotel which lies in a radius of 2 km from the given location ).

Using these two factors, you can prepare a list of hotels whose location and description match the user's needs.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.