Explainable-Recommendation

This is a project aimed to build a recommendation system to model user-item interaction. The recommendations are accompanied with predicted reviews, that a user might relate to.

Some examples of generated reviews: -

1. Predicted: -
   i is a great for my . and husband 's it and and the a minutes . and it was n't be happier . it was very sure how was a the , i was the tv , but it tv has he have it . far to install and tv to matter how you can using the tv . great buy . is . tv tv tv
   Actual: -
   this was a present for myself , my son installed it , in about 25 minutes , and i could n't be happier . i was n't sure this was for me as it for my bedroom , but my husband and i love it . so easy to move the tv no matter where you are in the room . a great buy .
   

2. Predicted: -
   i bought this on a sony cd and was a recognized the dvd player i , was to work me lot bit of i it cd is it a . i returned to it cd and . but was it of the the few . pen is is had with . is . . .
   Actual: -
   i tried this with a cd-rom player that was not recognizing the cd . initially it seemed to help a little bit . but the problem returned quite soon . so i tried using the cleaner again , i think three times in a row . the problem really never went away .
   

Dependencies

1. Pytorch
2. Spacy
3. Nltk
4. tqdm

References

1. Jianmo Ni, Julian McAuley. Personalized Review Generation by Expanding Phrases and Attending on Aspect-Aware Representations.
2. Neural Collaborative Filtering.