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

+ Recent posts