Dierent proposals have been made in recent years to exploit
Social Web tagging data to improve recommender systems.
The tagging data was used for example to identify similar
users or viewed as additional information about the recommendable
items.
In this work we propose to use tags as a means to express
which features of an item users particularly like or dislike.
Users would therefore not only add tags to an item but also
attach a preference or rating to the tag itself, expressing, for
example, whether or not they liked a certain actor in a given
movie. Since rating data is in general sparse in commercial
recommender applications we also present how to infer the
user opinion regarding a certain feature (tag) for a given
item automatically. In contrast to previous works, we not
only infer the user's general preference for a tag but rather
determine this preference in the context of a certain item.
An evaluation on the MovieLens data set reveals that
our new tag-enhanced recommendation algorithm is slightly
more accurate than a recent tag-based recommender even
when the explicit tag rating data is 100% sparse, that is, if
only derived information can be used.
%0 Journal Article
%1 gedikli2010rating
%A Gedikli, Fatih
%A Jannach, Dietmar
%D 2010
%J Systems and the Social Web at ACM
%K collaborative rating recommendation tag
%T Rating items by rating tags
%X Dierent proposals have been made in recent years to exploit
Social Web tagging data to improve recommender systems.
The tagging data was used for example to identify similar
users or viewed as additional information about the recommendable
items.
In this work we propose to use tags as a means to express
which features of an item users particularly like or dislike.
Users would therefore not only add tags to an item but also
attach a preference or rating to the tag itself, expressing, for
example, whether or not they liked a certain actor in a given
movie. Since rating data is in general sparse in commercial
recommender applications we also present how to infer the
user opinion regarding a certain feature (tag) for a given
item automatically. In contrast to previous works, we not
only infer the user's general preference for a tag but rather
determine this preference in the context of a certain item.
An evaluation on the MovieLens data set reveals that
our new tag-enhanced recommendation algorithm is slightly
more accurate than a recent tag-based recommender even
when the explicit tag rating data is 100% sparse, that is, if
only derived information can be used.
@article{gedikli2010rating,
abstract = {Dierent proposals have been made in recent years to exploit
Social Web tagging data to improve recommender systems.
The tagging data was used for example to identify similar
users or viewed as additional information about the recommendable
items.
In this work we propose to use tags as a means to express
which features of an item users particularly like or dislike.
Users would therefore not only add tags to an item but also
attach a preference or rating to the tag itself, expressing, for
example, whether or not they liked a certain actor in a given
movie. Since rating data is in general sparse in commercial
recommender applications we also present how to infer the
user opinion regarding a certain feature (tag) for a given
item automatically. In contrast to previous works, we not
only infer the user's general preference for a tag but rather
determine this preference in the context of a certain item.
An evaluation on the MovieLens data set reveals that
our new tag-enhanced recommendation algorithm is slightly
more accurate than a recent tag-based recommender even
when the explicit tag rating data is 100% sparse, that is, if
only derived information can be used.},
added-at = {2014-03-01T06:49:23.000+0100},
author = {Gedikli, Fatih and Jannach, Dietmar},
biburl = {https://www.bibsonomy.org/bibtex/2e7380137d10bd6a765897ea54bd05a31/inmantang},
interhash = {7a4e1b28558c54b576678146c5a614fe},
intrahash = {e7380137d10bd6a765897ea54bd05a31},
journal = {Systems and the Social Web at ACM },
keywords = {collaborative rating recommendation tag},
timestamp = {2014-03-01T06:49:23.000+0100},
title = {Rating items by rating tags},
year = 2010
}