# [expertfinder-dev] Contributions from the Information Retrieval field

Gianluca Demartini demartini at L3S.de
Mon Nov 19 14:33:04 GMT 2007

Hi all,

FYI, a lot of systems and formal models have been proposed in the IR =

field so far.
Please find attached herewith a brief overview of the current =

approaches: it does not consider the formal models proposed in the IR =

community; if you are interested I can point you to some other works.

Bests,
Gianluca

-- =

Gianluca Demartini, MSc
PhD Student at L3S Research Center
Leibniz Universit=E4t Hannover

Room 240, 2nd floor, Appelstr. 4, DE-30167 Hannover, Germany
Phone: +49 (0)511 762 17756
Fax: +49 (0)511 762 17779
Mobile: +39 349 5119466
http://www.l3s.de/~demartini

-------------- next part --------------
\section{Related Work}\label{sec:relwork}
%ES systems
The topic of Expert Search (ES) is a relatively new one but
already several systems have been proposed. The systems proposed
in the past use several information and features like
Social Network information;
co-occurrences of terms and
changes in the competencies of the people;
rule-based models and FOAF\footnote{http://www.foaf-project.org/}
data; and using post on Web Forums \cite{www07}.
One of the first approaches is the =

Enterprise PeopleFinder \cite{mclean2003epc}
also known as P at noptic Expert \cite{craswell2001pan}.
This system first builds a candidate profile
attaching all documents related to her/him,
giving
different weights to the documents based on their
type (e.g. an homepage is more important than other
web pages), in
one big document which represents the candidate.
The problem of this systems is that it can only consider
the terms of the documents as topics of expertise and
that the candidate name matching (i.e. the name appears
into the document or not) and the relationship
between candidate and documents is only binary (i.e. =

a document is related to the candidate or it is not).
=

A common idea in the community of ES is that two different
approaches are possible: \emph{Expert finding} and =

\emph{Expert profiling} \cite{balo:sear06}. The main idea is that the form=
er
approach aims at retrieving first the documents relevant
to the query and  extract the experts from them. The latter
approach first builds a profile for each candidate and then
match the query with the profiles without considering the documents
anymore \cite{balog2007det}.	=

%ES models
Some steps in
the direction of defining formal models for ES have been made and models b=
ased on
probability and language models \cite{Azzopardi2006, balog2006fme,
balog2006fea} have been proposed and are widely used \cite{trec2006}.
Another work which proposes a model for ES  	=

as a voting problem. In this model the documents associated
with the candidate are viewed as votes for the candidate expertise.
In this approach they use the Retrieval
Status Value of the documents  but the problem is that =

the relationship between candidates and documents is not
weighted and a document can only be related to a candidate
or not.
=

=

=

@inproceedings{craswell2001pan,
title=3D{{P at noptic Expert: Searching for Experts not just for Documents}},
author=3D{Craswell, N. and Hawking, D. and Vercoustre, A. and Wilkins, P.},
booktitle=3D{Ausweb, 2001},
year=3D{2001},
url =3D "http://es.ciro.au/pubs/craswell_ausweb01.pdf" =

}
@Article{trec2006,
author =3D {},
title =3D {The Fifteenth Text REtrieval Conference(TREC 2006) Proceedings},
year  =3D{2006},
journal =3D {NIST Special Publication: SP 500-266}
}

@article{mclean2003epc,
title=3D{{Enterprise PeopleFinder: Combining Evidence from Web Pages and =
Corporate Data}},
author=3D{McLean, A. and Vercoustre, A.M. and Wu, M.},
journal=3D{Proc. Australian Document Computing Symposium},
year=3D{2003}
}

@inProceedings{balo:sear06,
author =3D    {K. Balog and de Rijke, M.},
title =3D     {Searching for People in the Personal Work Space},
booktitle =3D {International Workshop on Intelligent Information Access =
(IIIA-2006)},
year =3D      2006,
}

@inproceedings{Azzopardi2006,
title =3D {Language Modeling Approaches for Enterprise Tasks},
author =3D {Azzopardi, L. and Balog, K. and  de Rijke, M.},
booktitle =3D {The Fourteenth Text Retrieval Conference (TREC 2005)}
=

}
@article{balog2006fme,
title=3D{{Formal models for expert finding in enterprise corpora}},
author=3D{Balog, K. and Azzopardi, L. and de Rijke, M.},
journal=3D{Proceedings of the 29th annual international ACM SIGIR confere=
nce on Research and development in information retrieval},
pages=3D{43--50},
year=3D{2006},
publisher=3D{ACM Press New York, NY, USA}
}

@article{balog2006fea,
title=3D{{Finding experts and their Details in e-mail corpora}},
author=3D{Balog, K. and de Rijke, M.},
journal=3D{Proceedings of the 15th international conference on World Wide=
Web},
pages=3D{1035--1036},
year=3D{2006},
publisher=3D{ACM Press New York, NY, USA}
}

@article{balog2007det,
title=3D{{Determining Expert Profiles (With an Application to Expert Find=
ing)}},
author=3D{Balog, K. and de Rijke, M.},
journal=3D{Proceedings of IJCAI-2007},
year=3D{2007}
}

@article{macdonald2006vot,
title=3D{{Voting for Candidates: Adapting Data Fusion Techniques for an E=
author=3D{Macdonald, C. and Ounis, I.},
journal=3D{Proceedings of the ACM Conference on Information and Knowledge=
Management (CIKM) 2006},
=

year=3D{2006}
=

}

@article{www07,
title=3D{Expertise Networks in Online Communities: Structure and Algorith=
ms},