As there is a large amout of textual information in research articles as compared to other items like camra, movies, accessories and songs which don't have that much information rather abstract level meta data is available for such items. which out of CB and CF approaches is best suited for this domain?
What does your 'CB' stand for? (content-based? but content-based WHAT?)
My short answer about the general challenge facing recommender systems is, that the context always matters -- and encoding the context appropriately will make your recommender system generally better, and will help reduce both false positives and false negatives (but from my experience, will esp. help w.r.t. false positives). Now, all these other examples have their contexts, but those contexts are generally different; and for each of those application domain, there is some amount of meta-data (which however may or may not be used by a recommender system). So I don't think that recommending research articles is necessarily fundamentally all that different, from other typos of domain-specific reco. systems. So to me, the main question is, how to adequately and relatively succinctly capture and encapsulate the most pertinent meta-data/contextual data, to make recommendations both efficient and accurate?
Any of the two approaches would do fine depending on what you want to achieve. In User-User Collaborative filtering , your recommender system tries to consider the similarities between each user and other users on the online platform and based on these similarities, research articles that closer neighbors of the target user showed interest in are equally recommended to him. These approach gives room for more diverse or serendipitous recommendations. On the other hand, in Content-based approach, recommendations are made to the user based on his research interest profile. Research articles are only recommended to the user if he had shown interest in that topic or keywords previously. This approach lacks some diversity and serendipity but may recommend research articles that the user may like now just because he had liked it before.
But it should be noted that research interest is time-dependent and as such our recommender should consider that as well. This link below may guide you further:
Article Modelling a Collaborative Digital Research Ecosystem Based o...
Dear Predrag Tosic,
By CB we mean Content-based recommendation approach for recommending items. there are two main types of recommendation one is content-based and other is Collaborative filtering. in Content based approach item features and meta data related to items is used to recommend items matching user profile. on the other hand, in CF user friend opinions are used to recommend items. here like-minded users play vital role. system first find user friends and then on the bases of these friends items are recommended consequently .
If you have more ratings, CF is better. Otherwise, CBF is better.
- Badges
- Science topic