• Martin Leginus
  • Valdas Žemaitis
2. term, Computer Science, Master (Master Programme)
Social tagging facilitates users to annotate and categorize
items (Web links, pictures, movies, etc.).
Assigned tags express the user preferences. Social
tagging systems (STS) have become popular hence
various tag based recommenders have been developed.
The state-of-the-art STS model three types of entities
(i.e. tag-user-item) and relationships between them
are encoded into a 3-order tensor. Latent relationships
and patterns can be discovered by applying
tensor factorization techniques like Higher Order
Singular Value Decomposition (HOSVD), Canonical
Decomposition (CD), etc. STS accumulate large
amount of data that signicantly slows down the
process of a tensor factorization. Firstly, we propose
to reduce tag space by exploiting clustering
techniques so that execution time is improved and
memory requirements are decreased while preserving
the quality of the recommendations. The clustering is
motivated by the fact that many tags in a tag space
are semantically similar thus the tags can be grouped.
Secondly, we propose to incorporate the personal prior
knowledge to increase the precision of tensor based
recommenders. In addition, we take an advantage
of Non-negative Tensor Factorization (NTF) to get
rid of negative values from the factorized tensor that
are dicult to interpret. Finally, we combine all the
approaches to improve the quality and time of computations
and present the promising experimental results.
Publication date31 May 2011
Number of pages30
ID: 52685298