Notes on a Possible Partnership  of the Department of Electronics  and
Computer Science (DEI) of the University of  Padova (Italy) within the
FET Proposal  "Supporting   Collective   Intelligence  on  the    Web"
(http://www.cpm.mmu.ac.uk/~bruce/bsi/FET-project.html)

Massimo Melucci
melo@dei.unipd.it



The FET  proposal describes a  software  system on  the Web aiming  to
implement a mechanism of self-organization and adaptation of Web sites
supporting the exchange of knowledge among researchers.

The system will be based on a shared semantic network which will adapt
itself  to the  needs of  its  users on the   basis of the active  and
collaborative use of the network itself. This means, for example, that
nodes independently  produced by different users  can be clustered and
linked together on the basis of the their content.

The semantic network  will be adapted  using algorithms being designed
to add, cluster, rank,  or remove nodes  or links of the network.  The
adaptation will be performed  automatically after an initial period of
time during which  users will interact  using  classical Web tools  to
build   an  initial knowledge.   This    period  of  time is    called
"bootstrapping". Once    an  initial  configuration   has been  built,
automatic procedures will adapt it.

The Information Management Systems research group (IMS) has been doing
research on associative networks   since late eighties.  Specifically,
its research members has actively involved  in the research addressing
Information  Retrieval (IR), hypertexts   for IR, automatic  hypertext
construction.

The partnership of  DEI, and  specifically of  IMS  can be consist  in
participating  to the following main  points described in the document
"Boostrapping Social Intelligence"
(http://www.cpm.mmu.ac.uk/~bruce/bsi/bsi.html)


*An associative network*

Weights or priorities of links can  be computed using IR techniques as
done for the automatic  construction  of hypertexts. These  techniques
aims to build links  among nodes on the  basis of the semantic content
of the nodes themselves.

These techniques can  then complement those  based on  the topology of
links and on the pattern of link traversal. The same techniques can be
used   to cluster together  similar questions  and  similar answer, to
rank, and then to decide if nodes  and links are promoted, downgraded,
or eliminated [1].


*A shared semantic network*

The  IMS research group has  recently  worked on  semantic networks as
natural generalization of associative networks. Algorithms employed to
automatically construct associative networks can be used to assist the
users on  creating a shared network where  both nodes  and link labels
have to be managed. A specific competence can  for example be spent on
hypertext for IR [3].


*Database update*

The presence of  a persistent  data  management system  is a necessary
condition  so that the  whole system  can work. Indeed,  we  can speak
about adaptation only if  we do not need to  rebuild the whole network
from scratch. The system can be implemented with a database management
system, perhaps a relational one, or with  a IR system, accordingly to
the emphasis is placed  on  consistency, or on content  representation
respectively. The system can layered on two  or three levels where the
client level is implemented through a Web browser, and the data server
level can be implemented through a database system [2].


*References*


[1] M. Agosti, F. Crestani, and M. Melucci.  Design and implementation
     of   a  tool for the    automatic  construction of  hypertexts for
     Information Retrieval.  Information Processing   &     Management,
     32(4):459--476, July 1996.

[2]   M. Agosti,  L. Benfante,   and M.   Melucci.  OFAHIR: on-the-fly
     automatic authoring of  hypertexts for information  retrieval.  In
     S. Spaccapietra  and   F. Maryanski,  editors,   "Data Mining  and
     Reverse  Engineering:   Searching   for Semantics",   IFIP,  pages
     269--300. Chapman and Hall, 1998.

[3] M.  Agosti, L.  Benfante, F.  Crestani,  V.  Del Re,  M.  Lisanti,
     G.   Mazzini,  A.    Schisano.    Study   of  Semantic    Networks
     Inter-Operations: Final Report.    Vitrociset Space Division   and
     Consorzio   Padova  Ricerche, I.   VCS/CPR/FR/1.0, 30  March 1999,
     pp.140