Presentazione di PowerPoint

1
Search Computing Stefano Ceri
2
Who helped me in doing the proposal
3
Candidate ideas

• Web Engineering
• Software Engineering for the Semantic Web
• Stream reasoning
• Search computing

4
How I picked the SeCo idea.
5
Time to delivery of the proposal

• Paraboschi Thursday evening
• Rinaldi Friday morning (in the shower)
• Week-end Main Body Written
• Monday-Wednesday State-of-the-art, Burocracy
• Thursday First submission
• Friday_at_5PM Deadline

6
Motivating Examples

• Who are the strongest candidates in Europe for
  competing on software ideas?

• Who is the best doctor who can cure insomnia in
  a close-by hospital?

• Where can I attend an interesting scientific
  conference in my field and at the same time relax
  on a beautiful beach nearby?

This information is available on Internet, but no
software system is capable of computing the
answer. Queries span over multiple semantic
domains and require composing ranking of results.
7
Their Common Aspect

• Multi-domain queries
• The answers are on the Web

• A knowledgeable user would do the query
  step-by-step
• Search database conferences, get their city
• Check that the city average temperature is warm
  enough
• Search low-cost flights via a broker for that
  city
• Search luxury hotels via another broker

• After hours of painful search the user might
  actually succeed!
• Can this be done better?

8
Genesis of Search Computing

• 2003 LOWELL WORKSHOP CHALLENGE Find an ethnical
  restaurant in a nice place close to Milano .

• Logically a composition of domains
• Restaurants (ethnical)
• Geo-locations (nice place close to Milano)

• Composing maps with geo-located information is
  now almost solved by many services, i.e. on top
  of yahoo local, google-local

9
Camorra-Controlled Locations near Naples
10
Genesis of Search Computing

• 2003 LOWELL WORKSHOP CHALLENGE Find an ethnical
  restaurant in a nice place close to Santa Clara

• Logically a composition of domains
• Restaurants (ethnical)
• Geo-locations (nice place close to Santa Clara)

• Composing maps with geo-located information is
  now almost solved by many services, i.e. on top
  of yahoo local, google-local

• Composing maps with geo-located information is
  now almost solved by many services, i.e. on top
  of yahoo local, google-local

• but in general no system is capable of
  composing arbitrary semantic domains

11
Search Computing

• Search computing is a new multi-disciplinary
  science which will provide the abstractions,
  foundations, methods, and tools required to give
  answer multi-domain queries on the Web

• Emphasis on
• Search services. These are software services
  producing ranked information. Ranking is
  essential for search service composition.

• Search Integration. The objective is not to build
  new search systems but instead to integrate a
  world-wide network of search systems.

• Web site http//home.dei.polimi.it/ceri/seco/

12
Research Chapters of Search Computing, 1

• Foundational theories, rooted into formal
  disciplines such as mathematics and optimization
  theory.
• Statistical models for estimating the number and
  qualities of the results produced by a search
  service.
• Optimization methods for determining efficient
  plans for service integration
• Software paradigms for designing and constructing
  search computing systems.
• Interaction paradigms to help user-friendly
  expression of queries and ranking.
• Framework for search-oriented software
  architectures and their instrumentation.

13
Research Chapters of Search Computing, 2

• Semantic domain knowledge for dealing with
  terminological aspects in composing search engine
  results.
• Higher-order rankings for prioritizing search
  objects and services.
• Personal and social aspects for setting ranking
  in relationship to individuals and context.
• Business models for pushing and developing search
  computing economy.
• Legal and privacy issues concerned with search
  sources integration.
• Advanced computational architectures for search
  computing.

14
Some Challenging Issues

• Definition of best multi-facet,
  multi-disciplinary problem cutting across the
  various areas
• understanding and translation of encoding systems
  for ranking,
• contextualization relative to the situation in
  which the query is presented,
• knowledge about the individual who is asking the
  question,
• societal behaviours such as the presence in
  preferred lists or in authoritative sources, such
  as blogs.
• Dynamic ranking problem dynamically ranking and
  selecting search objects and services at query
  execution time
• Includes object-based ranking,
• Driven by user interaction.

15
How does it relate to

• Semantic Web?
• At least in the first part of SeCo we define and
  solve a simpler problem, where sources are
  pre-defined and their semantics is know and
  therefore join patterns are given.
• Dynamic selection will be slowly and carefully
  injected.
• Social Web?
• We use uses societal information to improve the
  search, selection, and ranking of objects and
  sources. But we dont plan to do specific
  research, rather to reuse results.

16
How does it relates to other projects?

• Pharos (large IP with little DEI involvment)
• Deals with extraction of metadata from new media
  (video, audio files) and their Internet search
  (find all videos where Obama speaks about Italy)
• It is orthogonal, but we may use early results.
• larKC (large IP with DEI medium involvment)
• Deals with developing new languages, methods and
  tools for massive reasoning we are involved in
  stream reasoning.
• Streams are yet another form of ranked
  information (by their timestamps) where recent
  data are more valuable, so we may reuse some of
  the research approaches and methods.
• Bio-informatics
• An important scientific domain with lots of open
  issues

17
Preliminary Results in Search Computing

• FUNDING (2007-08) PRIN NGS (New Generation
  Search)
• Politecnico Milano (National Coordination)
• University of Roma 3
• Free University of Bolzano
• The brick join of two search services
• Information Systems, March 2008
• The framework multi-domain query optimization
• International Very Large Data Bases Conference,
• Auckland (NZ), August 2008
• The interface mash-up based interaction
• IEEE-Internet Computing, November 2008

18
JOIN of Web Services

• Input items resulting from TWO web service
  calls, possibly ranked
• Output composed items resulting from the
  concatenation of matching items, presented in a
  global ranking order
• Matching condition using
• value equality,
• partial set matching
• term matching within a vocabulary
• ..
• Services are known, their matching function is
  predefined this is not service discovery!

19
Popupular rock CDs

• Sources Amazon and iTunes

• Score measures ranking
• Match measures similarity

20
Relevant news in two newspapers

• Sources Corriere.it and Repubblica.it

21
Search Service Integration Framework

• Objective a Web Service Management System
• The system accepts queries, optimize them
  transparently to the user, and produces the
  result
• This is the follow-up of research done at
  Stanford VLDB06 but with significant changes
• Focus on search services
• Ranking as first-class citizen
• Physical optimization

22
Example
Find database conferences in the next six months
in warm locations offering inexpensive flights
from Milano and luxury hotels
23
Query Plan
Composition of four services Conferences, Weater
forecasts, Flights, Hotels
24
Method
Approach
25
Developer-Oriented interface

• Mashing up software services is becoming very
  popular among developers
• We propose a declarative mashup language for
  search services as a simple interface of the Web
  Service Management System, hiding all the
  optimization

26
Mashup interface for Search Service Integration
27
Some of the wrapped Web sites in the current
prototype

• Booking.com (www.booking.com) for hotels
• Expedia (www.expedia.it) for flights
• AccuWeather (http//www.dapper.net/) for weather
  conditions
• TicketOne (www.ticketone.it) for events
• GoogleMaps (maps.google.com) Distance Calculator
  Find Businesses
• Bed-and-breakfasts(www.bedandbreakfast.it)
• 35mm.it (http//programmazione.35mm.it/) for
  movies locations
• IMDB(www.imdb.com) for movies descriptions

28
Workplan Details
29
Research Agenda for 2009

• Building a theory of search computing focus on
  services producing ranked lists
• Operations, parameters, statistical models,
  optimization models
• Improving language abstractions and description
  formalisms for supporting multi-domain queries
• Covering various levels of abstractions
• Requiring various levels of users expertise
• Build a first prototype so far we have some of
  its ingredients
• Invest on applications more experiments, in
  diverse fields (including bio-medical
  applications)
• Start thinking to many other topics and
  organize the team.

30
Participants

• DB Group
• Daniele Maria Braga
• Marco Brambilla
• Alessandro Campi
• Stefano Ceri
• Sara Comai
• Emanuele Della Valle
• Piero Fraternali
• Pier Luca Lanzi
• Marco Masseroli
• Maristella Matera
• Giuseppe Pozzi
• Marco Tagliasacchi
• DEI
• Edoardo Amaldi
• DIG
• Roberto Verganti
• Tommaso Buganza

• Phd Students
• Adnan Adib
• Mamoun Abu Helou
• Davide Barbieri
• Alessandro Bozzon
• Davide Mazza
• Stefania Ronchi
• Massimo Tisi
• Post-Docs
• Michael Grossniklaus
• Davide Martinenghi
• Anthony Ventresque (?)

31
Advisory Board

• Fabio Casati, Università di Trento, expert on
  "Service Computing"
• Georg Gottlob, Oxford University, expert on
  "Search Computing Theory"
• Ioana Manolescu, INRIA - Paris, expert on
  "Systems and Performance"
• Roberto Verganti, School of Management,
  Politecnico di Milano, expert on "Business
  Models"
• Gerhard Weikum, Max-Planck-Institut für
  Informatik, Saarbruecken, expert on "Information
  Retrieval for the Web"
• Jennifer Widom, Stanford University, expert on
  "Languages and Paradigms"

32
Continuing.

• Davide Martinenghi Theory of Search Computing
• Alex Campi Join of Search Services
• Daniele Braga Framework for Search Computing
• Piero Fraternali Search for Video-Audio Contents
  
• Then A round table of first impressions among
  SeCo participants (expecially newcomers)
• Finally (for the ones who will deserve it) Big
  Dinner _at_ Ceris