Context Based Recommendation System Using Co-Attention Neural Network

  • B.Jothi , M.Pushpalatha,S.Krishnaveni,Vishesh N Jain


Context-based recommender systems include context as a parameter to retrieve recommendations with better accuracy. The primary purpose of a context-based recommender system is to form relationships between the user, item and its context to enhance the prediction. A lot of the information regarding the context is obtained from the data spread all across the internet. User reviews are a major part of information harnessing, providing information about the user’s needs, the item’s characteristics and the situation within which the review was written. It is an important source of data collection that can be used to enhance the user, item and context embedding. Recent networks that use contexts as one of the parameters only characterize relations between users and context, or items and contexts, which might not be enough to maximize accuracy, as the prediction is related to the users, items and contexts. This paper highlights a neural network that considers three entities instead of just two, called Context-based Co-attention Neural Network (CCANN) that actively retrieves relations between user, items and contexts which are then used to assess a co-attention score, which is the similarity between the user’s preferences and the item’s aspects up to a certain degree. Our main objective is to understand the trends when given different contextual information and to identify suitable outcomes for the user based on their preferences. To simplify our work further, we use an embedding method to concatenate and learn embedding of different entities as a singular unit using reviews as our source of information. This is to be tested on a dataset from Data Infiniti Hotel reviews, to further add to the observations that this architecture does indeed take acceptable time and has good accuracy.

How to Cite
B.Jothi , M.Pushpalatha,S.Krishnaveni,Vishesh N Jain. (2020). Context Based Recommendation System Using Co-Attention Neural Network. International Journal of Advanced Science and Technology, 29(9s), 690 - 698. Retrieved from