Distributed Information Retrieval and Applications

Fabio Crestani and Ilya Markov

University of Lugano, Via G. Buffi 13, 6900, Lugano, Switzerland

{[Link],[Link]}@[Link]

Abstract. Distributed Information Retrieval (DIR) is a generic area of research

that brings together techniques, such as resource selection and results aggrega-
tion, dealing with data that, for organizational or technical reasons, cannot be
managed centrally. Existing and potential applications of DIR methods vary from
blog retrieval to aggregated search and from multimedia and multilingual retrieval
to distributed Web search. In this tutorial we briefly discuss main DIR phases, that
are resource description, resource selection, results merging and results presen-
tation. The main focus is made on applications of DIR techniques: blog, expert
and desktop search, aggregated search and personal meta-search, multimedia and
multilingual retrieval. We also discuss a number of potential applications of DIR
techniques, such as distributed Web search, enterprise search and aggregated mo-
bile search.

1 Introduction
Distributed Information Retrieval (DIR), also known as Federated Search or Feder-
ated IR, concerns with aggregating multiple searchable sources of information under
a single interface [6,23]. DIR consists of the following phases: (i) resource descrip-
tion or representation, where a high-level description is built for each federated source.
(ii) Server/resource selection, where, given a user’s query, several relevant sources are
selected for further processing. (iii) Results merging or aggregation, where the results
obtained from selected sources are combined into a single result list. (iv) Results pre-
sentation, where the obtained results are grouped and positioned on a result page.
However, modern applications of the standard DIR techniques usually have a dif-
ferent set of assumptions and limitations [29]. For instance, aggregated search works
in highly cooperative environments and does not need to merge results into a single
list, blog and expert search do not require description and merging phases, while mul-
tilingual retrieval requires additional steps of query and document translation. In this
tutorial we discuss the standard DIR techniques and show how they can be adapted and
applied to various IR problems, such as blog distillation and desktop search, aggregated
search and distributed Web search, multimedia and multilingual retrieval.

2 Distributed Information Retrieval

In this section we briefly discuss main DIR phases, such as resource description, re-
source selection, results merging and results presentation.
Resource Description. In the offline phase a high-level description is built for each
federated source. The description may include a full content of a source (or only a sam-
ple of its documents in uncooperative environments [7]), term and document statistics,

P. Serdyukov et al. (Eds.): ECIR 2013, LNCS 7814, pp. 865–868, 2013.
c Springer-Verlag Berlin Heidelberg 2013
866 F. Crestani and I. Markov

metadata (if available) and other descriptors of the source’s content. The descriptions of
all federated sources are managed centrally by a DIR broker and are used for subsequent
phases, such as resource selection and results merging.
Resource Selection. Given a user’s query and the sources’ descriptions, the DIR bro-
ker selects the most relevant sources for the query. First generation resource selection
techniques, also known as large document approaches, represent each source as a con-
catenation of its documents. The obtained large documents are ranked using standard
IR techniques, such as adapted INQUERY in CORI [5] and language modeling in [33].
Second generation or small document approaches use a centralized sample index of
documents and rank sources based on the number and the position of their documents
in a centralized ranking (eg. ReDDE [26], CRCS [22] and others [15,18,20,32]). Finally,
classification-based resource selection combines the above approaches and a number of
other query- and corpus-based features in a machine learning framework [1,12].
Results Merging and Score Normalization. The user’s query is forwarded to the se-
lected sources and the retrieved source-specific results are merged into a single list us-
ing results merging and score normalization methods. Results merging techniques use
sources’ descriptions either implicitly through resource selection, like CORI [5,17], or
explicitly like SSL [25] and SAFE [24]. Score normalization methods do not use de-
scriptions, but require document relevance scores to be provided by federated sources [16].
Results Presentation. Instead of merging results into a single list or in addition to that,
the results may be presented to a user in various ways, eg. blended, tabbed, side-by-side,
etc. [28,31].

3 Applications
In this section we discuss the applications and adaptations of DIR techniques to various
IR tasks.
Applications of Resource Selection. Resource selection is probably the most widely
used DIR technique. For example, in blog distillation the most relevant blogs need to
be retrieved for a user’s query. Since each blog is a collection of posts, blog distilla-
tion can be treated as a resource selection problem and both small and large document
approaches can be applied to solve it [11,21]. The same idea is applicable to expert
search if each expert is considered as a collection of documents that he/she produced.
In desktop search each document type may be treated as a separate documents source
and resource selection may be performed to select the most relevant one [13].
Aggregated Search. Aggregated search augments Web search results with the results
of several vertical searches [3]. It can be seen as a variation of DIR, where aggregated
verticals are managed by the same content provider and, therefore, are fully coopera-
tive. Moreover, verticals are highly heterogeneous in content and types of media. Still
most of DIR steps are present here. First, resource description may be performed for ef-
ficiency reasons [3]. Second, vertical selection can be seen as a variation of resource se-
lection that accounts for multiple media types and where no vertical can be selected [3].
Finally, the vertical-specific results need to be blended into the Web results [2].
Other Applications. Multilingual retrieval combines documents written in multiple lan-
guages and, therefore, benefits from results merging methods [27]. Multimedia distributed
 Distributed Information Retrieval and Applications 867

digital libraries can use a standard DIR architecture [8]. A personal meta-search system
provides a user with a possibility of searching the Web and personal content with a single
interface using DIR techniques [30]. Similarly, federated Web search provides a low-cost
solution for aggregating multiple searchable Web sites within a single interface [19].

4 Potential Applications
In this section we discuss distributed Web search, enterprise search and aggregated
search on mobile devices as potential applications of DIR methods.
Distributed Web Search. Site selection in distributed Web search is a specific resource
selection problem, where search sites are not autonomous, resources and algorithms are
homogeneous and the distribution of content is managed explicitly [9]. Currently site
selection is performed based on a predicted query performance [4] or a thresholding
algorithm [10]. Although some work has been done on the feasibility of DIR methods
for topically partitioned collections [14], the applicability of existing resource selection
techniques to distributed Web search and the development of novel specific methods
still needs to be addressed.
Enterprise Search. Enterprise search involves different types of document sources
used within a company as well as a number of external searchable collections that need
to be aggregated into a single company-wide search system. DIR techniques such as
resource selection and results presentation may help in solving this problem [29].
Aggregated Mobile Search. Aggregated search on mobile devices has a number of
unique features, such as a rich context, a specific interaction of a user with search
results and technical limitations of mobile devices. The applicability of existing DIR
techniques and the development of novel task-specific methods for aggregated mobile
search are yet to be addressed.

References
1. Arguello, J., Callan, J., Diaz, F.: Classification-based resource selection. In: Proceedings of
CIKM, pp. 1277–1286. ACM (2009)
2. Arguello, J., Diaz, F., Callan, J.: Learning to aggregate vertical results into web search results.
In: Proceedings of CIKM, pp. 201–210 (2011)
3. Arguello, J., Diaz, F., Callan, J., Crespo, J.F.: Sources of evidence for vertical selection. In:
Proceedings of SIGIR, pp. 315–322 (2009)
4. Baeza-Yates, R., Murdock, V., Hauff, C.: Efficiency trade-offs in two-tier web search sys-
tems. In: Proceedings of SIGIR, pp. 163–170 (2009)
5. Callan, J.P., Lu, Z., Croft, W.B.: Searching distributed collections with inference networks.
In: Proceedings of SIGIR, pp. 21–28 (1995)
6. Callan, J.: Advances in Information Retrieval. In: Distributed Information Retrieval, vol. ch.
5, pp. 127–150. Kluwer Academic Publishers (2000)
7. Callan, J., Connell, M.: Query-based sampling of text databases. ACM Transactions of In-
formation Systems 19(2), 97–130 (2001)
8. Callan, J., Crestani, F., Nottelmann, H., Pala, P., Shou, X.M.: Resource selection and data fu-
sion in multimedia distributed digital libraries. In: Proceedings of SIGIR, pp. 363–364 (2003)
9. Cambazoglu, B.B., Plachouras, V., Baeza-Yates, R.: Quantifying performance and quality
gains in distributed web search engines. In: Proceedings of SIGIR, pp. 411–418 (2009)
10. Cambazoglu, B.B., Varol, E., Kayaaslan, E., Aykanat, C., Baeza-Yates, R.: Query forwarding
in geographically distributed search engines. In: Proceedings of SIGIR, pp. 90–97 (2010)
868 F. Crestani and I. Markov

11. Elsas, J.L., Arguello, J., Callan, J., Carbonell, J.G.: Retrieval and feedback models for blog
feed search. In: Proceedings of SIGIR, pp. 347–354 (2008)
12. Hong, D., Si, L., Bracke, P., Witt, M., Juchcinski, T.: A joint probabilistic classification model
for resource selection. In: Proceedings of SIGIR, pp. 98–105 (2010)
13. Kim, J., Croft, W.B.: Ranking using multiple document types in desktop search. In: Proceed-
ings of SIGIR. pp. 50–57 (2010)
14. Kulkarni, A., Callan, J.: Document allocation policies for selective searching of distributed
indexes. In: Proceedings of CIKM, pp. 449–458 (2010)
15. Markov, I.: Modeling document scores for distributed information retrieval. In: Proceedings
of SIGIR, pp. 1321–1322 (2011)
16. Markov, I., Arampatzis, A., Crestani, F.: Unsupervised linear score normalization revisited.
In: Proceedings of SIGIR, pp. 1161–1162 (2012)
17. Markov, I., Arampatzis, A., Crestani, F.: On CORI results merging. In: Serdyukov, P.,
Braslavski, P., Kuznetsov, S.O., Kamps, J., Agichtein, S.R.E., Segalovich, I., Yilmaz, E.
(eds.) ECIR 2013. LNCS, vol. 7814, pp. 736–739. Springer, Heidelberg (2013)
18. Markov, I., Azzopardi, L., Crestani, F.: Reducing the uncertainty in resource selection. In:
Serdyukov, P., Braslavski, P., Kuznetsov, S.O., Kamps, J., Agichtein, S.R.E., Segalovich, I.,
Yilmaz, E. (eds.) ECIR 2013. LNCS, vol. 7814, pp. 500–511. Springer, Heidelberg (2013)
19. Nguyen, D., Demeester, T., Trieschnigg, D., Hiemstra, D.: Federated search in the wild: the
combined power of over a hundred search engines. In: Proceedings of CIKM, pp. 1874–1878
(2012)
20. Paltoglou, G., Salampasis, M., Satratzemi, M.: Integral based source selection for uncooper-
ative distributed information retrieval environments. In: Proceedings of the ACM LSDS-IR
Workshop, pp. 67–74 (2008)
21. Seo, J., Croft, W.B.: Blog site search using resource selection. In: Proceedings of CIKM, pp.
1053–1062 (2008)
22. Shokouhi, M.: Central-Rank-Based Collection Selection in Uncooperative Distributed In-
formation Retrieval. In: Amati, G., Carpineto, C., Romano, G. (eds.) ECIR 2007. LNCS,
vol. 4425, pp. 160–172. Springer, Heidelberg (2007)
23. Shokouhi, M., Si, L.: Federated search. Foundations and Trends in Information Retrieval 5,
1–102 (2011)
24. Shokouhi, M., Zobel, J.: Robust result merging using sample-based score estimates. ACM
Transactions of Information Systems 27(3), 1–29 (2009)
25. Si, L., Callan, J.: Using sampled data and regression to merge search engine results. In:
Proceedings of SIGIR, pp. 19–26 (2002)
26. Si, L., Callan, J.: Relevant document distribution estimation method for resource selection.
In: Proceedings of SIGIR, pp. 298–305 (2003)
27. Si, L., Callan, J., Cetintas, S., Yuan, H.: An effective and efficient results merging strat-
egy for multilingual information retrieval in federated search environments. Information Re-
trieval 11(1), 1–24 (2008)
28. Sushmita, S., Joho, H., Lalmas, M., Villa, R.: Factors affecting click-through behavior in
aggregated search interfaces. In: Proceedings of CIKM, pp. 519–528 (2010)
29. Thomas, P.: To what problem is distributed information retrieval the solution? Journal of the
American Society for Information Science and Technology 63(7), 1471–1476 (2012)
30. Thomas, P., Hawking, D.: Server selection methods in personal metasearch: a comparative
empirical study. Information Retrieval 12(5), 581–604 (2009)
31. Thomas, P., Noack, K., Paris, C.: Evaluating interfaces for government metasearch. In: Pro-
ceedings of IIiX, pp. 65–74 (2010)
32. Thomas, P., Shokouhi, M.: Sushi: scoring scaled samples for server selection. In: Proceedings
of SIGIR, pp. 419–426 (2009)
33. Xu, J., Croft, W.B.: Cluster-based language models for distributed retrieval. In: Proceedings
of SIGIR, pp. 254–261 (1999)