Recommender System with Social Annotations using Neural Net Advancements
- 0 Collaborators
The study leverages from the recent advancements of Neural Networks in the field of recommendation modeling. Here we purpose to use the social annotations to gain more information about users and instead of user-item matrix, create a more sophisticated model with user-item vs other features (like for movie database features could be director, actor, producer etc.) and see if we can predict the product popularity based on this information. ...learn more
Project status: Under Development
Internet of Things, Artificial Intelligence
Intel Technologies
AI DevCloud / Xeon
Overview / Usage
Recommender systems assist users in navigating complex information spaces and focus their attention on the content most relevant to their needs. Often these systems rely on user activity or descriptions of the content. Social annotation systems, in which users collaboratively assign tags to items, provide another means to capture information about users and items. Each of these data sources provides unique benefits, capturing different relationships. We propose leveraging multiple sources of data: ratings data as users report their affinity toward an item, tagging data as users assign annotations to items, and item data collected from an online database. Taken together, these datasets provide the opportunity to learn rich distributed representations by exploiting recent advances in neural network architectures.
Methodology / Approach
Project data is collected from www.metacritic.com. The website aggregates the reviews of media products: movies, TV shows, music albums and video games. The data was collected by web-scrapping using python.
Two metrics are used to evaluate model performance: 1) RMSE and 2) NDCG
Models used until now for recommendation building: User-based collaborative filtering, Item-based collaborative filtering. Word2Vec is used to learn the context of related words in multi-dimensional vector space.
Next step is using deep learning techniques skip-gram-NS/CBOW to develop models and compare learnings.
Think of even better techniques if possible.
-Observe the trends in product popularity in future based on the user comments that the product page have so far. Or think of something which may make these trends clear.
Technologies Used
Recommender Systems using kNN
Word2Vec for NLP/ Neural Network techniques
Timeseries - (not yet used)
libraries used: Surpriselib,gensim, pandas, numpy, other custom libraries
Intel AI DevCloud architecture - used for large datasets since calculating the recommendations for these is a computationally expensive and requires high performing platform.