=Paper=
{{Paper
|id=Vol-497/paper-12
|storemode=property
|title=Tag Recommendation Using Probabilistic Topic Models
|pdfUrl=https://ceur-ws.org/Vol-497/paper_06.pdf
|volume=Vol-497
|dblpUrl=https://dblp.org/rec/conf/pkdd/KrestelF09
}}
==Tag Recommendation Using Probabilistic Topic Models==
https://ceur-ws.org/Vol-497/paper_06.pdf
Tag Recommendation using
Probabilistic Topic Models
Ralf Krestel and Peter Fankhauser
L3S Research Center
Leibniz Universität
Hannover, Germany
Abstract. Tagging systems have become major infrastructures on the
Web. They allow users to create tags that annotate and categorize con-
tent and share them with other users, very helpful in particular for
searching multimedia content. However, as tagging is not constrained by
a controlled vocabulary and annotation guidelines, tags tend to be noisy
and sparse. Especially new resources annotated by only a few users have
often rather idiosyncratic tags that do not reflect a common perspec-
tive useful for search. In this paper we introduce an approach based on
Latent Dirichlet Allocation (LDA) for recommending tags of resources.
Resources annotated by many users and thus equipped with a fairly sta-
ble and complete tag set are used to elicit latent topics represented as a
mixture of description tokens and tags. Based on this, new resources are
mapped to latent topics based on their content in order to recommend
the most likely tags from the latent topics. We evaluate recall and pre-
cision for the bibsonomy benchmark provided within the ECML PKDD
Discovery Challenge 2009.
1 Introduction
Tagging systems [1] like Flickr, Last.fm, Delicious or Bibsonomy have become
major infrastructures on the Web. These systems allow users to create and man-
age tags to annotate and categorize content. In social tagging systems like Deli-
cious the user can not only annotate his own content but also content of others.
The service offered by these systems is twofold: They allow users to publish
content and to search for content, thus tagging also serves two purposes for the
user:
1. Tags help to organize and manage own content, and
2. Find relevant content shared by other users.
Tag recommendation can focus on one of the two aspects. Personalized tag
recommendation helps individual users to annotate their content in order to
manage and retrieve their own resources. Collective tag recommendation aims
at making resources more visible to other users by recommending tags that
facilitate browsing and search.
However, since tags are not restricted to a certain vocabulary, users can pick
any tags they like to describe resources. Thus, these tags can be inconsistent
and idiosyncratic, both due to users’ personal terminology as well as due to the
different purposes tags fulfill [2]. This reduces the usefulness of tags in particular
for resources annotated by only a few users (aka cold start problem in tagging),
whereas for popular resources collaborative tagging typically saturates at some
point, i.e., the rate of new descriptive tags quickly decreases with the number of
users annotating a resource [3].
The main goal of the approach presented in this paper is to overcome the cold
start problem for tagging new resources. To this end, we use Latent Dirichlet
Allocation (LDA) to elicit latent topics from resources with a fairly stable and
complete tag set. The latent topics are represented as a mixture of description
tokens like URL, title, and other metadata, and tags, which typically co-occur.
Based on this, new resources are mapped to latent topics based on their descrip-
tion in order to recommend the most likely tags from the latent topics.
The remainder of this paper is organized as follows. In Section 2, we define
the problem of tag recommendation more formally, and introduce the approach
based on LDA. In Section 3 we present our evaluation results. In Section 4 we
discuss related work, and in Section 5 we summarize and outline possible future
research directions.
2 Tag Recommendation
2.1 Problem Definition
Given a set of resources R, tags T , and users U , the ternary relation X ⊆
R × T × U represents the user specific assignment of tags to resources. T consists
of two disjoint sets Ttag and Tdesc . Ttag contains all user assigned tags, Tdesc
contains the vocabulary of content and meta information, such as abstract or
resource description, which is represented as tag assignment by a special “user”.
A post b(ri , uj ) for resource ri ∈ R and a user uj ∈ U comprises all tags assigned
by uj to ri : b(ri , uj ) = πt σri ,uj X 1 . The goal of collective tag recommendation is
to suggest tags to a user uj for a resource ri based on tag assignments to other
resources by other users collected in Y = σr6=ri ∨u6=uj πr,t X ⊆ R × T .
2.2 Latent Dirichlet Allocation
The general idea of Latent Dirichlet Allocation (LDA) is based on the hypothesis
that a person writing a document has certain topics in mind. To write about
a topic then means to pick a word with a certain probability from the pool of
words of that topic. A whole document can then be represented as a mixture
of different topics. When the author of a document is one person, these topics
reflect the person’s view of a document and her particular vocabulary. In the
context of tagging systems where multiple users are annotating resources, the
1
projection π and selection σ operate on multisets without removing duplicate tuples
resulting topics reflect a collaborative shared view of the document and the tags
of the topics reflect a common vocabulary to describe the document.
More generally, LDA helps to explain the similarity of data by grouping fea-
tures of this data into unobserved sets. A mixture of these sets then constitutes
the observable data. The method was first introduced by Blei, et. al. [4] and ap-
plied to solve various tasks including topic identification [5], entity resolution [6],
and Web spam classification [7].
The modeling process of LDA can be described as finding a mixture of topics
for each resource, i.e., P (z | d), with each topic described by terms following
another probability distribution, i.e., P (t | z). This can be formalized as
N
X
P (ti | d) = P (ti |zi = j)P (zi = j | d), (1)
j=1
where P (ti ) is the probability of the ith term for a given document and zi is
the latent topic. P (ti |zi = j) is the probability of ti within topic j. P (zi = j) is
the probability of picking a term from topic j in the document. These probability
distributions are specified by LDA using Dirichlet distributions. The number of
latent topics N has to be defined in advance and allows to adjust the degree of
specialization of the latent topics. The algorithm has to estimate the parameters
of an LDA model from an unlabeled corpus of documents given the two Dirichlet
priors and a fixed number of topics. Gibbs sampling [5] is one possible approach
to this end: It iterates multiple times over each tag t, and samples a new topic
j for the tag based on the probability P (zi = j|t, z−i ), where z−i represents all
topic-word and document-topic assignments except the current assignment zi
for tag t, until the LDA model parameters converge.
Application to Tagging Systems LDA assigns to each document latent top-
ics together with a probability value that each topic contributes to the overall
document. For tagging systems the documents are resources r ∈ R, and each
resource in addition to its description from Tdesc is described by tags t ∈ Ttag
assigned by users u ∈ U . Instead of documents composed of terms, we have
resources composed of tags. To build an LDA model we need resources and asso-
ciated tags previously assigned by users. For each resource r we need some posts
b(r, ui ) assigned by users ui , i ∈ {1 . . . n}. Note that for each resource, at least
the tag assignments from its description is available. Then we can represent each
resource in the system not with its actual tags but with the tags from topics
discovered by LDA.
For a new resource rnew with few or no posts, we can expand the latent
topic representation of this resource with the top tags of each latent topic. To
accomodate the fact of some tags being added by multiple users whereas others
are only added by one or two users we can use the probabilities that LDA assigns.
As formalized in Equation 1 this is a two level process. Probabilities are assigned
not only to the latent topics for a single resource but also to each tag within a
latent topic to indicate the probability of this tag being part of that particular
Table 1. Top terms composing the latent topic “images” and “tutorial”
Tag Count Prob. Tag Count Prob.
images(tag) 243 0.064 tutorial(tag) 640 0.185
photo(tag) 218 0.057 howto(tag) 484 0.140
photography(tag) 205 0.054 tutorial(desc) 204 0.059
image(tag) 188 0.049 tutorials(tag) 184 0.053
photos(tag) 164 0.043 tutorials(desc) 173 0.050
photo(desc) 138 0.036 tips(tag) 126 0.037
images(desc) 106 0.028 reference(tag) 118 0.034
photos(desc) 98 0.026 guide(tag) 79 0.023
flickr(tag) 93 0.024 lessons(tag) 50 0.014
pictures(desc) 61 0.016 tips(desc) 48 0.014
graphics(tag) 49 0.013 wschools(desc) 45 0.013
media(tag) 48 0.013 tutoriel(tag) 33 0.010
art(tag) 48 0.013 comment(tag) 29 0.008
topic. We represent each resource ri as the probabilities P (zj |ri ) for each latent
topic zj ∈ Z. Every topic zj is represented as the probabilities P (tn |zj ) for
each tag tn ∈ T . By combining these two probabilities for each tag for rnew ,
we get a probability value for each tag that can be interpreted similarly as the
tag frequency of a resource. Setting a threshold allows to adjust the number of
recommended tags and emphasis can be shifted from recall to precision.
Imagine a resource with the following tags: “photo”, “photography”, and
“howto”. Table 1 shows the top terms for two topics related with the assigned
tags. The latent topics comprise a broad notion of (digital) photography and the
various aspects of tutorial material. Given these topics we can easily extend the
current tag set or recommend new tags to users by looking at the latent topics.
If LDA assumes that our resource in question belongs to 66% to the “photo”-
topic and to 33% to the “howto”-topic, these probabilities are multiplied with
the individual topic/tag probabilities, and the top five tags recommended are
“tutorial”, “howto”, “images”, “photo”, and “photography”.
3 Evaluation
We used the data provided by the ECML PKDD Discovery Challenge 2009 to
evaluate our approach and fine-tune our parameters. For assessing precision,
recall, and f-measure we used the supplied evaluation script.
3.1 Dataset
Our dataset consists of the provided training data for the Discovery Challenge.
All experiments were performed on the post-core at Level 2, where all tags, users,
and resources occur at least in two posts. To measure the performance of our
system, we split the training data into a 90% training set and a 10% test set
based on posts (called content IDs in the dataset).
Table 2. Fields parsed to represent a resource
Bibtex Bookmark
Author Title URL
Editor Description Description
Booktitle Journal Extended
Abstract
Table 3. Actual tags and recommended tags with computed probablity for URL
http://jo.irisson.free.fr/bstdatabase/
Real Tag LDA Tag LDA Prob.
latex bibtex(tag) 0.017
bibtex latex(tag) 0.017
bibliography bibtex(desc) 0.014
database latex(desc) 0.008
engine theory(desc) 0.005
style citeulike(desc) 0.005
tex bibliography(tag) 0.004
reference database(tag) 0.003
academic styles(desc) 0.003
For each resource, as defined by the hash values, we build up a textual repre-
sentation. This representation contains all the tags that were assigned by users
in the training set to a particular resource. In addition, we add terms extracted
from the description of the resource. More precisely, we tokenized different fields
describing a bookmark or bibtex entry. An overview of the fields can be seen in
Table 2. Afterwards, we removed stopwords and punctuation marks. Using also
the description ensures that we have some terms related to a resource even if no
other user before tagged it.
3.2 Results
The tag recommendation algorithm is implemented in Java. We used Mallet [8],
which provides an efficient SparseLDA implementation [9], to perform the Latent
Dirichlet Allocation with Gibbs sampling. The LDA algorithm takes three input
parameters: the number of terms to represent a latent topic, the number of latent
topics to represent a document, and the overall number of latent topics to be
identified in the given corpus.
Table 3 shows the actual tag distribution for a randomly selected resource
(http://jo.irisson.free.fr/bstdatabase/), the top tags recommended by
LDA with aggregated probabilities, and all the tags provided by a sample user.
As the actual tags indicate, the url is a database/latex related site. The tags
recommended by LDA come from six latent topics, comprising latex, databases,
academia, references, bibliography, and style. These tags characterize the re-
source quite well.
Table 4. F-measure for different number of recommended tags and different
number of LDA topics compared with recommending the most frequent tags
(mf)
# LDA topics
No. Tags
50 100 200 400 600 800 1000 2500 5000 10000 mf
1 0.170 0.191 0.214 0.229 0.229 0.230 0.229 0.238 0.240 0.235 0.270
2 0.200 0.225 0.248 0.266 0.271 0.271 0.274 0.289 0.288 0.283 0.335
3 0.209 0.233 0.257 0.277 0.282 0.285 0.287 0.302 0.303 0.300 0.362
4 0.209 0.237 0.257 0.279 0.287 0.289 0.292 0.305 0.307 0.303 0.379
5 0.209 0.238 0.258 0.280 0.286 0.291 0.293 0.307 0.307 0.304 0.388
Table 4 compares the f-measure reached for various numbers of latent topics
and the baseline which simply recommends the top most frequent tags for each
resource (mf) 2 . As can be seen, the best f-measure for LDA is reached between
2500 and 5000 latent topics, but it does not reach the baseline by far. The main
reason for this seems to be that the average number of tags per resource is
just 10.3 (7.4 distinct tags). This is significantly smaller than the number of
(distinct) tokens in a full-text abstract or document, to which LDA has been
applied traditionally. Moreover, there are only about 2.8 posts per resource.
Thus, there is on the one hand too little co-occurrence evidence for eliciting
latent topics, on the other hand there is too little overlap between users on a
resource to effectively predict tags via the latent topics of a resource for a new
post.
However, to deal with resources that have only few tags associated it makes
sense to combine tag recommendations based on most frequent tags with tag
recommendations based on latent topics. With f req(t, r) the frequency of tag t
annotated for resource r, one estimate of the probability of tag t given resource
r is as follows:
f req(t, r)
P1 (t | r) = P (2)
ti ∈r f req(ti , r)
This estimate can be combined with the estimate P2 (t | r) via latent topics
in Equation 1 by means of a mixture:
P (t | r) = λP1 (t | r) + (1 − λ)P2 (t | r). (3)
Table 5 shows that this combination achieves consistently better recall and
precision than the individual approaches. The largest gain is achieved for the
first recommended tag. Similar accuracies are achieved when varying the mixture
parameter λ between 0.3 and 0.9, and for a number of latent topics ≥ 1000.
2
Unless stated explicitly otherwise, we recommend at least one tag and at most the
number of tags annotated to a resource
Table 5. Evaluation results for tag recommendation based on most frequent
tags, based on 5000 latent topics, and their combination with λ = 0.5.
Most Frequent Tags Latent Topics Combination
No. Tags
Recall Prec. F-Meas. Recall Prec. F-Meas. Recall Prec. F-Meas.
1 0.190 0.467 0.270 0.165 0.437 0.240 0.214 0.537 0.306
2 0.274 0.430 0.335 0.232 0.380 0.288 0.302 0.479 0.370
3 0.329 0.403 0.362 0.271 0.343 0.302 0.357 0.441 0.394
4 0.370 0.388 0.379 0.298 0.316 0.307 0.393 0.415 0.404
5 0.400 0.377 0.388 0.316 0.299 0.307 0.421 0.398 0.409
Table 6. Evaluation results DC09 challenge Task 1 based on most frequent tags,
based on 5000 latent topics, and their combination with λ = 0.5.
Most Frequent Tags Latent Topics Combination
No. Tags
Recall Prec. F-Meas. Recall Prec. F-Meas. Recall Prec. F-Meas.
1 0.010 0.032 0.015 0.045 0.158 0.070 0.049 0.169 0.076
2 0.018 0.031 0.022 0.073 0.131 0.094 0.078 0.140 0.100
3 0.022 0.029 0.025 0.092 0.114 0.102 0.099 0.122 0.110
4 0.026 0.028 0.027 0.094 0.112 0.103 0.102 0.120 0.111
5 0.028 0.027 0.028 0.096 0.112 0.103 0.105 0.120 0.112
3.3 Setups and Results for the Challenge Submission
We have submitted tag recommendations for Task 1 and Task 2 in the ECML
PKDD Discovery Challenge 2009. Task 1 aims at recommending tags for arbi-
trary users annotating a resource in 2009 based on tag assignments until 2008.
Thus the test data contain tags, resources, and users which are not available in
the training data. The topic models have been trained on the full dataset, com-
prising about 9.3 Mio tokens for 415 K resources. The test set consists of 43002
posts. Table 6 compares the results for 5000 latent topics with the results using
the most frequent tags, and the combination of the two approaches 3 . Because
for the posts in the test set there are only about 0.3 posts per resource in the
training set, recommending only the most frequent tags does not recommend
any tags for most of the resources. Consequently, recall and precision are signif-
icantly lower than for the approach based on latent topics. The combination of
the two approaches achieves slightly but consistently better recall and precision.
Task 2 operates on the post-core at Level 2, where all tags, users, and re-
sources occur at least twice in the training data, which comprises about 750 K
tokens for 22389 resources. The test set consists of 778 posts, for which there
exist on average 5.8 posts in the training set. Table 7 again compares the results
3
Our submission to the DC09 challenge was based on 2500 latent topics without
combination with most frequent tags, which achieved an F-measure of 0.098.
Table 7. Evaluation results DC09 challenge Task 2 based on most frequent tags,
based on 5000 latent topics, and their combination with λ = 0.5.
Most Frequent Tags Latent Topics Combination
No. Tags
Recall Prec. F-Meas. Recall Prec. F-Meas. Recall Prec. F-Meas.
1 0.147 0.411 0.216 0.133 0.404 0.200 0.156 0.450 0.232
2 0.223 0.341 0.270 0.204 0.326 0.251 0.252 0.386 0.305
3 0.284 0.305 0.294 0.258 0.281 0.269 0.313 0.339 0.326
4 0.325 0.275 0.298 0.298 0.251 0.272 0.352 0.300 0.324
5 0.357 0.256 0.298 0.319 0.224 0.263 0.386 0.276 0.322
Table 8. Evaluation results for DC09 challenge Task 2 for 5000 latent topics
without content
No. Tags Recall Precision F-Measure
1 0.128 0.362 0.189
2 0.191 0.293 0.232
3 0.236 0.254 0.245
4 0.267 0.225 0.244
5 0.299 0.207 0.245
for the two individual approaches and their combination 4 . As is to be expected,
recall and precision are much better than for Task 1, because there is more knowl-
edge available about the tagging practices of users. Like in our internal tests tag
recommendation based on most frequent tags outperforms the approach based
on LDA, and the combination outperforms the individual approaches.
Table 8 shows the results when only tags are used to elicit latent topics. Re-
call and precision are consistently lower. Thus taking into account the content of
resources leads to more effective latent topics for tag recommendation. However,
this does not hold for tag recommendation based on most frequent tags. Rec-
ommending the most frequent content terms or tags consistently leads to lower
precision and recall.
4 Related Work
Tag recommendation has received considerable interest in recent years. Most
work has focused on personalized tag recommendation, suggesting tags to the
user bookmarking a new resource: This is often done using collaborative fil-
tering, taking into account similarities between users, resources, and tags. [10]
introduces an approach to recommend tags for weblogs, based on similar weblogs
tagged by the same user. Chirita et al. [11] realize this idea for the personal desk-
top, recommending tags for web resources by retrieving and ranking tags from
4
Our submission to the DC09 challenge was based on 5000 topics without combination
with the most frequent tags and no limit on the number of recommended tags. This
achieved an F-measure of 0.258.
similar documents on the desktop. [12] aims at recommending a few descriptive
tags to users by rewarding co-occuring tags that have been assigned by the same
user, penalizing co-occuring tags that have been assigned by different users, and
boosting tags with high descriptiveness (TFIDF).
Sigurbjörnsson and van Zwol [13] also look at co-occurence of tags to rec-
ommend tags based on a user defined set of tags. The co-occuring tags are then
ranked and promoted based on e.g. descriptiveness. Jaeschke et al. [14] compare
two variants of collaborative filtering and Folkrank, a graph based algorithm for
personalized tag recommendation. For collaborative filtering, once the similarity
between users on tags, and once the similarity between users on resources is
used for recommendation. Folkrank uses random walk techniques on the user-
resource-tag (URT) graph based on the idea that popular users, resources, and
tags can reinforce each other. These algorithms take co-occurrence of tags into
account only indirectly, via the URT graph. Symeonidis et al. [15] employ dimen-
sionality reduction to personalized tag recommendation. Whereas [14] operate on
the URT graph directly, [15] use generalized techniques of SVD (Singular Value
Decomposition) for n-dimensional tensors. The 3 dimensional tensor correspond-
ing to the URT graph is unfolded into 3 matrices, which are reduced by means
of SVD individually, and combined again to arrive at a more dense URT tensor
approximating the original graph. Tag recommendation then suggests tags to
users, if their weight is above some threshold.
An interactive approach is presented in [16]. After the user enters a tag for a
new resource, the algorithm recommends tags based on co-occurence of tags for
resources which the user or others used together in the past. After each tag the
user assigns or selects, the set is narrowed down to make the tags more specific.
In [17], Shepitsen et al. propose a recommendation system based on hierarchical
clustering of the tag space. The recommended resources are identified using user
profiles and tag clusters to personalize the recommendation results. Note that
they use tag clusters to recommened resources whereas we use LDA topics, which
can be considered clusters, to recommend tags.
[3] introduce an approach to tag recommendation using association rules.
Resources are regarded as baskets consisting of tags, from which association
rules of the form T1 → T2 are mined. On this basis tags in T2 are recommended
whenever the resource contains all tags in T1 . A comparison of this approach
with the approach presented in this paper can be found in [18].
When content of resources is available, tag recommendation can also be ap-
proached as a classification problem, predicting tags from content. A recent
approach in this direction is presented in [19]. They cluster the document-term-
tag matrix after an approximate dimensionality reduction, and obtain a ranked
membership of tags to clusters. Tags for new resources are recommended by
classifying the resources into clusters, and ranking the cluster tags accordingly.
5 Conclusions and Future Work
In this paper we have presented and evaluated the use of Latent Dirichlet Alloca-
tion for collective tag recommendation. Using selected features from the content
of resources, tags, and users, we elicit latent topics that comprise typically co-
occuring tags and users. On this basis we can recommend tags for new users and
resources by mapping them to the latent topics and choosing the most likely
tags from the topics. The approach complements simple tag recommendation
based on most frequent tags especially for new resources with only few posts.
Consequently, combining tag recommendations based on latent topics with tag
recommendations based on most frequent tags outperforms the individual ap-
proaches.
For future work we want to investigate approaches that take into account
individual tagging practices for personalized tag recommendation.
Regarding data sets, we also want to experiment with datasets from different
domains, to check whether photo, video, or music tagging sites show different
system behavior influencing our algorithms. Another interesting direction we
want to follow is to apply LDA not only for tag recommendation but to employ
it in the context of recommending resources.
6 Acknowledgments
This work was supported in part by the EU project IST 45035 - Platform for
searcH of Audiovisual Resources across Online Spaces (PHAROS).
References
1. Marlow, C., Naaman, M., Boyd, D., Davis, M.: Ht06, tagging paper, taxonomy,
flickr, academic article, to read. In Wiil, U.K., Nürnberg, P.J., Rubart, J., eds.:
HYPERTEXT 2006, Proceedings of the 17th ACM Conference on Hypertext and
Hypermedia, August 22-25, 2006, Odense, Denmark, New York, NY, USA, ACM
(2006) 31–40
2. Golder, S., Huberman, B.A.: Usage patterns of collaborative tagging systems.
Journal of Information Science 32(2) (April 2006) 198–208
3. Heymann, P., Ramage, D., Garcia-Molina, H.: Social tag prediction. In: SIGIR ’08:
Proceedings of the 31st annual international ACM SIGIR conference on Research
and development in information retrieval, New York, NY, USA, ACM (2008) 531–
538
4. Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent dirichlet allocation. Journal of Machine
Learning Research 3 (January 2003) 993–1022
5. Griffiths, T.L., Steyvers, M.: Finding scientific topics. Proc Natl Acad Sci U S A
101 Suppl 1 (April 2004) 5228–5235
6. Bhattacharya, I., Getoor, L.: A latent dirichlet model for unsupervised entity
resolution. In: SIAM Conference on Data Mining (SDM). (April 2006) 47–58
7. Bı́ró, I., Siklósi, D., Szabó, J., Benczúr, A.A.: Linked latent dirichlet allocation in
web spam filtering. In: AIRWeb ’09: Proceedings of the 5th International Workshop
on Adversarial Information Retrieval on the Web, New York, NY, USA, ACM
(2009) 37–40
8. McCallum, A.K.: Mallet: A machine learning for language toolkit.
http://mallet.cs.umass.edu (2002)
9. Yao, L., Mimno, D., McCallum, A.: Efficient methods for topic model inference
on streaming document collections. In: KDD ’09: Proceedings of the 15th ACM
SIGKDD conference on Knowledge Discovery and Data Mining, New York, NY,
USA, ACM (2009)
10. Mishne, G.: Autotag: a collaborative approach to automated tag assignment for
weblog posts. In: WWW ’06: Proceedings of the 15th international conference on
World Wide Web, New York, NY, USA, ACM (2006) 953–954
11. Chirita, P.A., Costache, S., Nejdl, W., Handschuh, S.: P-tag: large scale automatic
generation of personalized annotation tags for the web. In: WWW ’07: Proceedings
of the 16th international conference on World Wide Web, New York, NY, USA,
ACM (2007) 845–854
12. Xu, Z., Fu, Y., Mao, J., Su, D.: Towards the semantic web: Collaborative tag
suggestions. In: Proceedings of Collaborative Web Tagging Workshop at 15th
International World Wide Web Conference. (2006)
13. Sigurbjörnsson, B., van Zwol, R.: Flickr tag recommendation based on collective
knowledge. In: WWW ’08: Proceeding of the 17th international conference on
World Wide Web, New York, NY, USA, ACM (2008) 327–336
14. Jäschke, R., Marinho, L.B., Hotho, A., Schmidt-Thieme, L., Stumme, G.: Tag
recommendations in folksonomies. In Kok, J.N., Koronacki, J., de Mántaras, R.L.,
Matwin, S., Mladenic, D., Skowron, A., eds.: Knowledge Discovery in Databases:
PKDD 2007, 11th European Conference on Principles and Practice of Knowledge
Discovery in Databases, Warsaw, Poland, September 17-21, 2007, Proceedings. Vol-
ume 4702 of Lecture Notes in Computer Science., Heidelberg, Germany, Springer
(2007) 506–514
15. Symeonidis, P., Nanopoulos, A., Manolopoulos, Y.: Tag recommendations based
on tensor dimensionality reduction. In: RecSys ’08: Proceedings of the 2008 ACM
conference on Recommender systems, New York, NY, USA, ACM (2008) 43–50
16. Garg, N., Weber, I.: Personalized, interactive tag recommendation for flickr. In:
RecSys ’08: Proceedings of the 2008 ACM conference on Recommender systems,
New York, NY, USA, ACM (2008) 67–74
17. Shepitsen, A., Gemmell, J., Mobasher, B., Burke, R.D.: Personalized recommen-
dation in social tagging systems using hierarchical clustering. In Pu, P., Bridge,
D.G., Mobasher, B., Ricci, F., eds.: Proceedings of the 2008 ACM Conference on
Recommender Systems, RecSys 2008, Lausanne, Switzerland, October 23-25, 2008,
New York, NY, USA, ACM (2008) 259–266
18. Krestel, R., Fankhauser, P., Nejdl, W.: Latent dirichlet allocation for tag recom-
mendation. In: 3rd ACM Conference on Recommender Systems, New York City,
USA, October 22-25, 2009, to appear. (2009)
19. Song, Y., Zhuang, Z., Li, H., Zhao, Q., Li, J., Lee, W.C., Giles, C.L.: Real-time
automatic tag recommendation. In: SIGIR ’08: Proceedings of the 31st annual
international ACM SIGIR conference on Research and development in information
retrieval, New York, NY, USA, ACM (2008) 515–522