Software Agents Knowledge Sharing KQML, KIF and Ontologies

1
Software AgentsKnowledge SharingKQML, KIF and
Ontologies

• Tim Finin
• University of Maryland
• Baltimore County
• finin_at_umbc.edu
• http//umbc.edu/finin

http//www.cs.umbc.edu/finin/sisce97
2
Overview

• Motivation
• Whats an agent?
• Agent theory
• Agent technology
• Applications
• Conclusion

Only when sage warriors use intelligent agents
will they certainly have success.
-- Sun Tzu
3
Motivation
4
About your workstation

• What work is your workstation doing for you
  today?
• Is it just sitting idle while you are here at
  this meeting and when you are asleep?
• Is it your workstation or anybodys ?
• Has it adapted to you and your preferences and
  behaviors?
• If your workstation cant do something, does it
  know how to find another which can?
• Can it collaborate and cooperate with its kin on
  the Internet?

5
About your web pages

• Why cant your web pages seek out viewers.
• When they change, can they notify the agents
  which are interested in them?
• What can your web pages say to be understood by
  machines?

6
About your web browser

• Why do you have to be in the drivers seat for it
  to work?
• Why can it understand any of the stuff that flows
  thorough it?

7
Whats an agent?
8
What is an agent?A brief tour of agent-space
An agent is a powerful and ubiquitous abstraction
in computer science

• Daemons (e.g., ftp agent)
• User interface clients (e.g., mail agent)
• Physical agents (e.g., robotics)
• Believable agents (e.g., VR and graphics)
• Intelligent software agents

9
What is a software agent?
We can identify some sub-categories

• User-facing agents
• Intelligent human-computer interface agents
• Adaptive user modeling agents
• Personal (expert) assistants
• e.g., personal retrieval agent
• Mobile software technology
• Cooperating software agents
• Resource discovery agents
• Mediators and facilitators
• Market agents
• etc

10
So, whats a software agent?

• There isnt a consensus yet, but several key
  concepts are important to this emerging paradigm.
• A software agent
• is an autonomous, goal-directed process
• is situated in, is aware of, and reacts to its
  environment
• cooperates with other agents
  (software or human) to accomplish its
  tasks

11
an emerging system-building paradigm
(with hype)
12
Why another paradigm?

• Because it addresses current issues
• Fully distributed and heterogeneous
• provides inherent scalability
• Proactive as well as reactive
• agents generalize client-server
• Adaptable and evolvable
• through interoperability and learning
• Easier for us to model and understand
• by taking an intentional stance
• Scales over many orders of magnitude
• Its agents, not turtles, all the way down

13
Software Agent Characteristics
after Hyacinth Nwana, 1996
14
Agent Characteristic Adaptation

• Agents adapt to their environment and users and
  learn from experience.
• Via machine learning, knowledge discovery,
  statistical techniques, etc.
• Via communication through exchange of metadata,
  brokering, and facilitation.
• Interface agents acquire and use user models
• Situated in and aware of their environment

15
Agent Characteristic Cooperation

• Agents use standard languages protocols to
  cooperate and collaborate to achieve common
  goals.
• Cooperate with human agents and other software
  agents
• Supported by agent communication languages and
  protocols.
• Consistent with human conventions and intuition.
• Toward team formation and agent ensembles

16
Agent Characteristic Autonomy

• Agents act autonomously to pursue their agenda.
• Proactive and reactive
• Goal directed behavior
• Appropriately long-lived
• Multi-threaded behavior
• Encourage viewing from an intentional stance

17
Agent Characteristic Mobility?
A mobile agent is an executing program that
migrates from machine to machine in a
heterogeneous network under its own control.

• Note -- this definition implies some agent
  attributes, e.g. autonomy, persistence, ...
• Mobile agents off some very interesting
  advantages as well as some disadvantages.
• This is an important technology for distributed
  systems and is largely orthogonal to other
  agent issues.

18
Agents vs. Intelligent Agents
What makes an agent an intelligent agent?

• A The size of the price tag.
• More seriously
• The paradigm covers agents of varying degrees of
  intelligence
• Intelligent agents will tend to
• know and apply more sophisticated domain
  knowledge
• recognizing underlying goals and intentions
• react to unexpected situations in a robust manner
• etc.
• Much of what we will be saying applies to agents
  of little or no intelligence.

19
Agent theory,architecturesand
ProgrammingTechnology
20
Agent Theory
There is a body of theory underlying agent-based
systems

• Knowledge representation and reasoning
• BDI models
• Speech acts
• Machine learning
• Game theory
• Econometric models

21
BDI agent models
Agents must represent and reason about other
agents, human or software.

• The BDI agent model considers agents to have
• Beliefs about itself, other agents and its
  environment
• Desires about future states (i.e., goals)
• Intentions about its own future actions (i.e.,
  plans)
• The BDI model is particularly useful for
• Developing formal models of agents
• Developing a deep model of agent communication
• Inferring an agents internal state from its
  behavior

22
Rational Agents

• A rational agents is modeled as making choices
  resulting in intentions in an attempt to
  optimize the expected utility with respect to
  their desires and consistent with their beliefs.
• Rational agents are self-interested agents
• A reasoner is limited in the amount of reasoning
  it can (or should) spend on any given problem.

23
Communication

• We draw on research in (computational)
  linguistics for an underlying communication model
• Speech act theory
• Discourse analysis
• We also draw on work from distributed computing
• Analyzing and specifying protocols

24
What is a speech act?
Speech act theory is a high level framework
developed by philosophers and linguists to
account for human communication

• Speakers do more than just utter sentences which
  have a logical (truth-theoretic) meaning.
• The cat is on the mat
• Speakers perform speech acts requests,
  suggestions, promises, warnings, threats,
  criticisms, praise, etc.
• I hereby promise to buy you lunch
• Every utterance is a speech act

25
Performatives

• In speech act theory, a performative is a speech
  act whose utterance performs an action.
• Examples promise, assert, request, confirm, ask
  ...
• Heuristic X is a speech act if you can say I
  hereby X.
• The KQML agent communication language uses the
  term performative to mean a primitive message
  type
• A key objective is to develop systems based on
  the right set of communication primitives

26
Econometric models
Economics has studied how to model, predict and
control the aggregate behavior of very large
collections of intelligent agents.

• Conceptual tools include
• game theory
• general equilibrium market mechanisms
• protocols for voting and auctions
• General assumption
• agents are only motivated by own self-interests
  and choose actions to maximize payoff given the
  current protocol.
• An objective is to design good artificial markets
  and protocols to result in the desired behavior
  of our agents

27
Artificial life

• One way that agents could adapt is via an
  evolutionary process.
• Individual agents use their own strategies for
  behavior with death as the punishment for poor
  performance and reproduction the reward for
  good.
• Techniques include
• genetic programming
• natural selection
• sexual reproduction

28
Agent Architectures
People are using several architectures for
agent-based information systems.

• Beyond client server
• Mediated architectures
• Single agents vs Multi-agent systems
• Markets and swarms

29
Beyond client-server
Agents generalize the client-server architecture
which has dominated the Internet since its
beginning.

• Think of your role in human society. Are you a
  client or a server?
• Mobile agents go beyond C/S by enabling clients
  to distribute computations to servers and
  vice-versa.
• Communicating agents generalize the C/S model
  since sometimes an agent acts as a client and
  sometimes as a server.

mediated architecture for information systems
-- after Gio Wiederhold
30
Agents vs. MAS

• Some research focuses on developing sophisticated
  individual agents with advanced capabilities and
  other on multi-agent systems (MAS)
• MAS research is particularly focused on
• cooperation and collaboration
• team and coalition formation
• information sharing among the team
• joint beliefs, goals and plans

31
Agent programming languages
Although many languages can be used for writing
software agents, there are useful characteristics

• Multi-threaded
• Good communication support
• Support for security (communication execution)
• Support for maintaining DBs and/or KBs
• Mobility?
• Good GUI support
• Good interoperability features to support
  mediated architectures and component-based
  programming

32
What are people using

• Mobile agents require a language that supports
  mobility. Some examples
• Languages designed to work in a sandbox Java,
  Safe-Tcl, Safe-Python,
• True mobile languages Telescript, Tacoma,
  Agent-Tcl,...
• Java-based frameworks for mobile languages
  Aglets, Voyager, ...
• Communicating agents can be written in your
  favorite programming language
• Java, C/C, Lisp, Prolog, Smalltalk, Tcl, Perl
• with a suitable agent-communication module

33
Example -- TKQML

• TKQML application adds KQML functionality to any
  Tcl/Tk shell.
• Provides simple route to high level communication
  among Tcl/Tk based agents.
• Incorporates tools for a facilitated environment
  (resource brokering).
• Scripts are registered dynamically for handling
  incoming messages.
• Current processing is interrupted by handler, and
  resumes when message handling is complete.

34
Agent Communication

• Agent cooperation requires communication
• Opinion Agent-agent communication is the key to
  realizing the potential of the agent paradigm,
  just as the development of human language was a
  key to the development of human potential
• We assume that agents use an Agent Communication
  Language or ACL to communication information and
  knowledge.
• Genesereth (CACM, 92) defined a software agent as
  any system which uses an ACL to exchange
  information.

35
Some ACLs

• Is CORBA an ACL?
• Knowledge sharing approach
• KQML, KIF, Ontologies
• FIPA
• Ad hock languages
• e.g., SGIs Open Agent Architecture language

36
Knowledge Sharing Technology
37
Knowledge Sharing Effort

• The KSE is a distributed research project
  involving over a dozen research groups aimed at
  developing techniques, methodologies and software
  tools for knowledge sharing and knowledge reuse.
• This requires a common language (syntax,
  semantics, pragmatics)
• Some existing components that can be used
  independently or together
• KIF -- knowledge interchange format (syntax)
• Ontolingua - a language for defining sharable
  ontologies (semantics)
• KQML - a high-level interaction language
  (pragmatics)

38
Knowledge Interchange Format
KIF

• KIF First order logic with set theory
• An interlingua for encoded knowledge
• Takes translation among n systems from O(n2) to
  O(n)
• Common language for reusable knowledge
• Implementation independent semantics
• Highly expressive -- can represent knowledge in
  typical application knowledge bases.
• The subject of an ANSI standardization committee
  (ASC NCITS.TC.T2)

39
KIF 101
KIF

• KIF is a relational language
• (salary 015-46-3946 72000)
• (supervisor ?p ?p)
• but not a flat one, as in relational databases
• (gt ( (width chip1) (length chip1))
• ( (width chip2) (length chip2)))
• KIF includes a variety of logical operators
• (gt (and (real-number ?x) (even-number ?n))
• (gt (expt ?x ?n) 0))

40
KIF 101
KIF

• KIF can encode meta-knowledge about functions
  relations
• (transitive R), (inverse R1 R2), (holds ?r 1 2) ,
  (value ?f 1 2)
• (ltgt (transitive ?r)
• (gt (and (holds ?r ?x ?y) (holds ?r ?y ?z))
• (holds ?r ?x ?z))))
• KIF can encode knowledge about knowledge,
• (interested joe (salary ,?x ,?y ,?z))
• KIF can also be used to describe procedures
• (progn (fresh-line t)
• (print "Hello!")
• (fresh-line t))
• Many agents only need
  or use a subset of
  KIF

41
KIF Software
KIF

• Several KIF-based reasoners are available from
  Stanford (Lisp implementation).
• IBMs ABE (Agent Building Environment) and RAISE
  reasoning engine use KIF as their external
  language.
• Stanfords Ontolingua uses KIF as its internal
  language.
• Translators (partial) exist for a number of other
  KR languages, including LOOM, Classic, CLIPS,
  Prolog, ...
• Parsers for KIF exist which take KIF strings into
  C or Java objects.
• See lthttp//www.cs.umbc.edu/kifgt for more
  information.

42
Common Semantics Shared Ontologies and Ontolingua
Ontologies

• Ontology A common vocabulary and agreed upon
  meanings to describe a subject domain.

Ontol"ogy (?), n. Gr. the things which exist
(pl.neut. of , , being, p.pr. of to be) -logy
cf.F. ontologie. That department of the science
of metaphysics which investigates and explains
the nature and essential properties and relations
of all beings, as such, or the principles and
causes of being. Webster's Revised Unabridged
Dictionary (G C. Merriam Co., 1913, edited by
Noah Porter)

• This is not a profoundly new idea
• Vocabulary specification
• Domain theory
• Conceptual schema (for a data base)
• Class-subclass taxonomy
• Object schema

43
Reusable Declarative Knowledge
Ontologies

• Declarative knowledge is the critical enabler for
• Large scale intelligent systems
• Interoperation of object-oriented systems
• Encoding knowledge is prohibitively expensive
• Extensive time
• Sophisticated expertise
• We must develop capability for knowledge reuse
  via tools and techniques for
• Creating and preserving reusable knowledge bases
• Assembling knowledge bases from reusable modules

44
Representation Requires Vocabulary
Ontologies

• Domain experts use domain-specific vocabulary
• Traditional KR languages are domain-independent
  (E.g, FOPC, frames, schemas) and dont provide a
  domain-specific vocabulary
• KB construction therefore involves two steps
• Defining vocabulary to be used to represent the
  knowledge
• Representing the knowledge using the defined
  vocabulary

45
Ontologies
Ontologies

• An ontology contains specification of concepts to
  be used for expressing knowledge
• -- Types of entities -- Relations and functions
• -- Attributes and properties -- Constraints
• Ontologies are distinguished from KBs not by
  their form, but by the role they play in
  representing knowledge
• consensus models for a domain
• Emphasis on properties that hold in all
  situations
• Emphasis on classes rather than instances
• Intended to support multiple tasks and methods
• Dont change during problem solving and are
  suited for compiling into tools
• Need to satisfy a community of use
• Emphasis on collaborative development
• Emphasis on translation to multiple logical
  formalisms
• Useful for education

46
Conceptual Schemas
Ontologies

• A conceptual schema specifies the intended
  meaning of concepts used in a data base

Data Base
Table price stockNo integer cost float
Data Base Schema
Auto Product Ontology
price(x, y) gt (x, y) auto_part(x)
stock_no(x) x
retail_price(x, y, Value-Inc)
magnitude(y, US_dollars) y
Product Ontology
Conceptual Schema
Units Measures Ontology
47
Ontology Library and Editing Tools
Ontologies
Ontolingua is a language for building,
publishing, and sharing ontologies.

• A web interface to a browser/editor at
  http//ontolingua.stanford.edu/ and mirror sites.
• Internal representation is KIF augmented with a
  frame language ontology
• enables OO structuring of ontologies
• Ontologies can be translated into other content
  languages, including KIF, LOOM, Prolog, CLIPS,
  etc.
• Provides user with fully cross-referenced HTML
  documents

48
KQML Knowledge Query and Manipulation Language
KQML

• KQML is a high-level, message-oriented,
  communication language and protocol for
  information exchange independent of content
  syntax and ontology.
• KQML is independent of
• the transport mechanism (e.g., tcp/ip, email,
  corba objects, IIOP, etc.)
• Independent of content language (e.g., KIF, SQL,
  STEP, Prolog, etc.)
• Independent of the ontology assumed by the
  content.
• KQML includes primitive message types of
  particular interest to building interesting agent
  architectures (e.g., for mediators, sharing
  intentions, etc.)

49
A KQML Message
KQML

• Represents a single speech act or performative
• e.g., ask, tell, reply, subscribe, achieve,
  monitor, ...
• with an associated semantics and protocol and a
  list of attribute/value pairs
• e.g., content, language, from, in-reply-to,
  ...

50
Semantics for tell(A,B,X)
KQML

• bel(A,X), A states to B that A believes X to be
  true (for A).
• Pre(A) bel(A,X) AND know(A,want(B,know(B,S)))
  where S may be bel(A,X) or NOT(bel(A,X))
• Pre(B) intend(B,know(B,S))
• Post(A) know(A,know(B,bel(A,X)))
• Post(B) know(B,bel(A,X))
• Completion know(B,bel(A,X))
• Comment The completion condition and
• postconditions hold unless a sorry or error
• suggests Bs inability to properly acknowledge
• the tell.

51
Some KQML Performatives
KQML

• Basic query performatives evaluate, ask-if,
  ask-in, ask-one, ask-all
• Multi-response query performatives
  stream-in, stream-all
• Response performatives reply, sorry, error
• Generic informational performatives tell,
  achieve, cancel, untell, unachieve
• Generator performatives standby, ready,
  next, rest, discard, generator
• Capability-definition performatives
  advertise, subscribe, monitor, import, export
• Networking performatives register,
  unregister, forward, broadcast, route

52
Simple Query Performatives
KQML
ask-all(P)
tell((p1 p2 p3...))
ask-all(P)
tell(p1)
tell(p2)
tell(p3)
Agent-based Information Retrieval
July 1997
53
KQML protocols
KQML
Broker
Recommend
Recruit
Register
Agent-based Information Retrieval
July 1997
54
Lockheed/UMBC KQML Software Architecture
KQML

• Router
• Content independent message router.
• One per agent - handles sending and asynchronous
  receiving of messages.
• KRILs
• Knowledge Representation Interface Library.
• Maps incoming KQML messages to application
  handlers
• Standard agent facilitate building MAS systems
• Agent name server
• Logger, Controller, Visualizer, ...
• Brokers ...
• Proxy agents ...
• Mediators ...

Agent-based Information Retrieval
July 1997
55
KQML Utility Agents
KQML

• We have developed a number of KQML speaking
  utility agents based on standard ontologies.
• Agent Name Server Logger
• Message Router Authenticator
• Broker Communication Visualizer
• Controller Proxy Agent

• These facilitate the development of agent-based
  systems.

Agent-based Information Retrieval
July 1997
56
KQML Utility Agents
57
KQML APIs and System Interfaces
KQML

• KQML APIs
• KATS (Loral/UMBC)
• KAPI (Lockheed/EIT/Stanford)
• MAGENTA (Stanford)
• LogicWare (Cryzalys)
• Toronto - Lisp and C
• JavaAgentTemplate (Stanford), Jackal (UMBC) --
  Java
• TKQML (UMBC) -- Tcl/Tk
• System interfaces
• Languages Lisp, C, C, Perl, Prolog, Tcl,
  CLIPS, Scheme, Java, ...
• KR languages LOOM, KRSL, ...
• Systems DBMSs, SIMS, ...

58
Applications
59
Agent-Based Information Retrieval (ABIR) Mind Map
Adaptive
Collaborative
Interface
EachMovie
Phoaks
Fab
RARE/Tunes
Firefly
ReferralWeb
GroupLens
SimilaritiesEngine
Morse
SiteSeer
MovieCritic
Yenta
ABIR
Large,
Proactive
Heterogeneous,
Distributed
Letizia
RemembranceAgent
CARROT
InfoSleuth
Push
Retsina
SAIRE
Backweb
Knowbot
Marimba
All-in-one
Pointcast
Fastfind
SIFT
Metacrawler
TopicAGENTs
Metasearch
Metasearch
Netbot Jango
Shopbots
Profusion
Shopbot
Savvysearch
WebCompass
60
Knowbot ABIR Systems

• A knowbot is an ABIR system that provides a
  single query language to access a variety of
  information sources
• It serves as a representative for the user (which
  demands program autonomy).
• Examples of knowbots MetaCrawler, SavvySearch,
  Shopbot, BargainFinder, Fido.
• Metacrawler
• Queries a variety of search engines in parallel.
• Provides a uniform user interface.
• Merges results from different engines.
• Downloads and scans pages if necessary.
• Passes ads through.

61
Collaborative Filtration Systems

• A collaborative filtration system makes
  recommendations to a person based on the
  preferences of similar users, e.g. recommending
  people (Yenta, ReferralWeb), products (Firefly,
  Tunes, EachMovie, Morse, MovieCritic), or
  readings (Wisewire, Firefly, Fab, Phoaks)
• Content-based recommendation retrieves other
  documents similar to those the user liked
  earlier.
• Collaborative recommendation retrieves documents
  liked by other people similar to the user.

62
Example Firefly

• Has been applied to music, movies, web pages,
  books, etc.
• Uses several neighbor sets to improve precision.
• Will recommend people who rate a given page
  highly, and pages that a giver person rates
  highly (e.g., My Yahoo).
• Sets cookies incessantly.
• Developed user passport

63
Example Remembrance Agent

• Indexes personal files and e-mail.
• Embedded in Emacs.
• When you perform a task, it automatically
  suggests relevant documents.
• Provides continuous associative recall.

64
Example Letizia

• Agent browses as you browse.
• People typically browse depth-first Letizia
  browses breadth-first.
• Uses a variety of heuristics to identify
  interesting pages.
• When an interesting page is identified, it is
  displayed in a separate browser window.

65
Semantic web markup with SHOE

• Simple HTML Ontology Extensions developed by UMCP
  PLUS Group
• Ontological mark-up tags allow the web page
  author to add semantic knowledge of the pages
  content.
• Pages and sub-regions are the ground objects in
  the world.
• Definitions for classes, relations, attributes,
  axioms, etc. are also defined on web pages.

66
URL agents

• URL agents are responsible for, and understand
  small groups of related web pages and communicate
  with other agents with similar pages
• Each URL has an agent which can think, speak, and
  act for it.

67
URL Agents speak for one or more related web
pages
Ontology Page
URLA2
KB
WWW

Ontology Page
URLA1
KB
Related Web Pages
68
Example
WWW text
Shoe tag
Prerequisite CMSC 202 ltRELATION
hasPrereq TOhttp//
....CMSC-202gt (tell to urla42 from
spider387 content (prereq
CS-341 CS-202) ontology
http///courseontology.html language
KIF)
KQML message sent to URL agent
69
Course Ontology
ltONTOLOGY course-ontology" VERSION2.5"gt
ltONTOLOGY-EXTENDS meeting-ontology"
VERSION"2.1" PREFIXmtg"
URL"http//onto.org/mtg/2.1/"gt ltONTDEF
CATEGORYcourse" ISAmtg.periodic"gt ltONTDEF
RELATIONfrequency" ARGScourse timeinterval"gt
ltONTDEF RELATIONhasPrereq" ARGScourse
course"gt ltONTDEF INFERENCEgt ltONTIF
RELATIONhasPrereq 1a 2bgt ltONTIF
RELATIONchainPrereq 1b 2cgt
ltONTTHEN RELATIONchainPrereq 1a
2cgtlt/ONTDEFgt lt/ONTOLOGYgt
70
urlAgent Prototype Architecture
URL agent
KQML API
Referenced Ontologies
CLIPS
Semantic information from local web pages in KQML
Web robot
KQML API
Web pages with semantic markup added
Markup added by hand
Markup added via SHOE enabled editor
Custom markup agent
Raw web page
Raw web page
Raw web page
71
Linneaus

• Problem What can we use as an ontology to
  characterize what a document is about?
• Solution use a pre-existing, naturally
  occurring ontology, e.g.,
• Yahoo hierarchy -- 150K nodes
• Newsgroup hierarchy -- 5K newsgroups
• Encyclopedia articles -- 10K articles
• Approach automatically classify the target
  document with respect to the ontology corpus with
  telltale.

72
CARROT Cooperating Agent-based Routing and
Retrieval of Text

• Agents provide access to different corpora, using
  existing IR engines
• Agents share metadata with Broker agents, which
  route queries and new documents to the right
  place(s).
• Two enabling technologies
• KQML agent communication language
• N-gram processing

73
CARROT Assumptions and Architecture

• Dynamic corpus - new documents may arrive at any
  time
• Corpus is physically partitioned among many
  back-end processors, each represented by an
  agent
• Each back-end handles 1-10 GB
• interact with local IR/DB engines
• generate metadata, to be shared with the
  broker(s)
• metadata for a set of documents is a compressed
  n-gram profile
• Brokers
• collect metadata from back-end agents
• use Telltale to manage these metadata corpora
• may be organized in hierarchies, thereby scaling
  to many GBs of data
• Light-weight version of Telltale to generate
  metadata for other engines

74
(No Transcript)
75
CARROT Back-end Agent

• Interacts with local IR/DB engines to access data
• No interference with existing applications
• Generates metadata, to be shared with one or more
  brokers
• Metadata for a set of documents is a compressed
  n-gram profile

76
CARROT Broker

• Collects metadata from back-end agents
• Uses Telltale to manage these metadata corpora
• Otherwise similar to back-end agents
• Can be organized in hierarchies

TK GUI
KQML
TKQML
broker.c
TCL Agent
Metadata
Telltale
77
Telltale

• Telltale is an information retrieval engine
  designed for
• scalability
• use with a wide variety of document types and
  languages
• embedding in larger systems
• Some key features include
• vector space model for representing documents and
  queries
• use of character n-grams
• use of corpus centroids as
  
  metadata
• Tcl/Tk user interface
• Agent api using KQML

78
Documents as term vectors

• A document is a sequence of terms (e.g. words,
  word stems or n-grams)
• We can represent a document as a (normalized)
  vector of the frequency of each of the unique
  terms in a document.
• A similarity measure of two documents can be
  measured as the angle between their vectors.
• We can represent the combination of two documents
  by adding (and renormalizing) their vectors
• We can represent a corpus as the sum of the
  vectors of all of its documents (the centroid)

79
n-grams vs. words

• An IR system can use n-grams or words as terms
• Advantages of n-grams
• Dont need a morphological model (e.g., for
  stemming) so good for multi-linguistic
  environment or non-language corpora (e.g., Java
  code).
• Robust w.r.t. letter errors (typos, ocr errors,
  etc)
• Provides some context since they span adjacent
  words
• computer science --gt compu ter_ s er_sc
  r_sci _scie ...
• Advantages of words
• Can be more precise (computation but not
  computer)
• Amenable to boolean combinations
• Bottom line?
• Depends on specifics of application

80
VR based visualization of retrieval

• Real-time, interactive stereo viewing of results
  of information retrieval engine
• Each document returned is rendered as a glyph
  (icon)
• Document properties mapped to 3D location, shape,
  color, transparency, and texture.
• Provides spatialization of complex relationships
  and comprehensible display of multiple variables

81
VR Approach

• Immersive
• Isolates the user from environment
• Expensive
• Minimally-immersive
• Access to environment
• Collaboration possible
• Low cost
• Two hands give proprioception
• Uses Two 3D Trackers with Buttons
• User manipulates 3D scene with trackers
• Each hand has a distinct role -- left sets up
  context and right performs fine manipulation

82
Visualizing a document space
Mappings X similarity to federal reserve
bank Y similarity to commodity prices Z
similarity to foreign exchange rate of the
dollar Shape similarity to coup attempt
against Noreiga with cube as lowest and cone as
highest Color age of document with blue as the
oldest and yellow as the newest Transparency Text
ure
83

• Participants
• IBM Corp
• University of Maryland Baltimore County
• University of North Carolina at Charlotte
• University of Florida
• Berclain USA Ltd.
• QAD Inc
• GSE Systems
• Lucent Technologies
• Ingersoll-Rand Co.
• Demand Solutions
• DLoG Remex Inc.
• Intercim
• EnvisionIt Software

• Funder National Institute of Standards and
  Technology / Advanced Technology Program
  (NIST/ATP)
• Technologies for the Integration of Manufacturing
  Applications (TIMA)
• 22.9M over 3 years (6/96 - 6/99), 45
  government, 55 consortium
• Goal Plug and Play framework of business
  objectives and integration enabling tools
  allowing a suite of solutions that can be
  implemented out-of-the-box at small and
  midsized manufacturing and process sites
  including MES, ERP, Finite Scheduling, and
  Capacity Analysis/Decision Support
• Objectives interoperability, configurability,
  adaptability, extensibility, plug and play.

84
Current Implementation Components
People
MES
Netscape
1
2
KQML/KIF
1
1
Java-Enabled Web Browser
BOD
3
4
4
4
MOOPI

• All agents register with the Agent Name Server
• An agent may communicate/collaborate with any
  other agent
• The Broker Agent matches agent interests with
  service providers
• Arrows indicate path of rate monitoring scenario

85
Conclusion
86
Some key ideas

• Software agents offer a new paradigm for very
  large scale distributed heterogeneous
  applications focusing on the interactions of
  autonomous, cooperating processes which can adapt
  to humans and other agents.
• Intelligence is always a desirable characteristic
  but is not strictly required by the paradigm.
• There is a wealth of prior theory from several
  disciplines but you neednt to be an expert to to
  apply or use it.

87
Some key ideas

• Modern programming languages support building
  agents with threads a real plus and mobility an
  interesting option.
• Communication is of central importance and, in
  particular
• Establishing common agent communication
  languages.
• Developing common ontologies for application
  domains.
• Establishing common agent protocols.
• The paradigm is still forming.
• See http//www.cs.umbc.edu/agents