Content-based approach
Its roots in information retrieval and information filtering
Focus on recommending items containing textual information, such as documents, Web sites (URLs), and news messages etc
Improvement by the use of user profiles that contain information about users’ tastes, preferences, and needs
The process of Content-based recommendation
Construction of per-user content-based profile
TF : Term Frequency, IDF : Inverse Document Frequency
N : Number of the documents
ni : How many times keyword ki is seen in the document
fi,j : Number of times keyword ki is seen in the document dj
Similarity measurement
u(c,s) : the utility function
ContentBasedProfile(c) = (wc1, …, wck) : the profile of user c
cosine similarity measure
Limitations of content-based approach
Limited Content Analysis
automatic feature extraction is much harder to multimedia data
cannot distinguish between a well-written article and a badly written one
Overspecialization
no experience, no recommendation
New User Problem