Software Agent

Chapter three:

Agent communication and knowledge sharing

Contents
 Agent communication and knowledge sharing.
 Introduction to Ontology Engineering
 KQML as an Agent Communication Language

08/08/2021 1
 Knowledge Level Communication

What is needed to communicate?

 Using the same terminology is essential!

 Assigning the same meaning to terminology is also essential!

 Having the same context is also essential!

 Shared terminology, context and meaning is needed

 Communication as well as representation language is essential

 Interaction as well as transport protocols are vital

08/08/2021 2
 Knowledge Level Communication

 Communication language: a medium through which the attitudes

regarding the content of the exchange are communicated

 A representation language(meaning, context and terminology)

 A communication language

08/08/2021 3
 Knowledge Level Communication

 Interaction protocol:
 knowing with whom to talk and how to find them

 knowing how to initiate and maintain an exchange

 Transport protocol: actual transport mechanism used for the

communication: HTTP, TCP, SOAP, etc.

 Communication as a social consensus

 Not only between two agents but a society of agents

08/08/2021 4
Knowledge Sharing

 Aims at defining a common language for message passing among

software modules in terms of inputs and outputs
 KQML and KIF as languages for information exchange and knowledge
level communication
 Ontolingua: an ontology description language and design tool

08/08/2021 5
Ontology: Definition /1 choose

 The American Heritage Dictionary defines ontology as the branch

of metaphysics that deals with the nature of being.

 The term ontology has been adopted by the AI community to refer

to a set of concepts or terms that can be used to describe some area
of knowledge or build a representation of it.

 Recently, the term ontology is usually used in conjunction with

knowledge sharing and reuse and it has been extensively used in
knowledge management (KM) research.
08/08/2021 6
What is Ontology? (1) list

 Types of AI theories
 Mechanism theories:
 “Secrets of making machines intelligent”

e.g., rule-based system, neural nets, fuzzy logic, unification, etc.)

 Content theories:
 A kind of theory of the domain on which the mechanism theory works

 Need a representation language

08/08/2021 7
What is Ontology? (2)

 Ontologies are content theories about the sorts of objects, properties

of the objects and relations between objects

 Ontologies provide potential terms of describing our knowledge

about the domain

 Predicate logic augmented with type-of relation (to induce class

hierarchies) can potentially be used to represent ontologies

08/08/2021 8
What is Ontology? (3) Describe

 Two Approaches

 Ontology as a representational vocabulary

 Conceptual structure of terms should remain unchanged during translation

 Ontology as the body of knowledge describing a domain

(commonsense domain in particular)

08/08/2021 9
What is Ontology? (4) Choose

 Philosophy viewpoint:
 Dealing with the nature of being or reality (Theory of
Existence)

 AI viewpoint:
 An explicit representation of conceptualization

 A theory of what entities can exist in the mind of a

knowledgeable agent

08/08/2021 10
What is Ontology? (5)

 A concept is a unit of thoughts consisting of two parts, the

extension and the intension.
 The extension covers all objects belonging to this concept and

 the intension comprises all features valid for all those objects

 Knowledge Base viewpoint

 A system of primitive vocabulary/concepts and their
relationships used for building artificial systems

08/08/2021 11
Ontology in KM Choose

 In knowledge management applications, ontologies are typically

used for three purposes:
 (1) to support knowledge visualization, where ontologies are inspected in order
to create new knowledge by analysis and recombination of existing knowledge
 (2) to support knowledge search, retrieval and personalization, where ontologies
are used to improve search and retrieval of information.
 (3) to serve as the basis for reasoning and reuse, where some degree of
formality of ontologies allows partial automation of problem solving and
reasoning.

08/08/2021 12
Ontology: Properties

 Ontologies have to deal with

 Heterogeneity in space: ontology as networks of meaning
 Tools for dealing with conflicting definitions and strong support in
interweaving local theories are essential in order to make this
technology workable and scalable

 Development in time: learning and adapting ontologies

 Having protocols for the process of evolving ontologies is a real
challenge. Evolving over time is an essential requirement for ontologies

08/08/2021 13
Why Ontology? (1)

 Knowledge sharing problem:

 Different systems use different concepts and terms for describing domains. This
makes it difficult to take knowledge out of one system and use it in another

 Ontologies may be used in an integration task to describe the semantics of

the information sources and to make the content explicit.

 Developing reusable ontologies that facilitates sharing and reuse is an

important goal of ontology research

 Developing tools supporting building, merging and translating between

ontologies is another goal

08/08/2021 14
Why Ontology? (2)

 Explicit ontology supports

 Shared understanding among agents
 Interoperability between tools
 Systems engineering
 Reusability
 Defining the vocabulary and concepts
 Plays the role of Mathematics in Physics (Theory of content)
 Defines the necessary and sufficient concepts and “methods”
 Common knowledge theory

08/08/2021 15
Ontology as Artifact

 A pragmatic view:

 An ontology is an artifact that must be constructed (i.e.,

designed artifact)

 Many artifacts have strong ontological flavour, such as

 Glossary, taxonomy, dictionary, thesaurus
 Knowledge bases, database schema
 Conceptual maps, templates
08/08/2021 16
Conceptual Graph:
DefinitionDescribe

 Conceptual Graph (CG) is a knowledge representation system

developed by John Sowa.

 CG is made of concept nodes and relation nodes

 concepts nodes represent entities, attributes, states, and events

 relations nodes show how the concepts are interconnected

08/08/2021 17
 Ontology vs. KB: Differences

 An ontology provides the basic structure around which a knowledge

base can be built

 An ontology provides a set of concepts and terms for describing

some domain and captures the intrinsic conceptual structure of the
domain and does not change with the change of environment

 This is done via Ontological Analysis

 Then, a knowledge base uses those terms to represent what is true

about some real or hypothetical world
08/08/2021 18
 Ontological Analysis (1)

 Ontological analysis clarifies the structure of knowledge. Given a domain, its

ontology forms the heart of any system of knowledge representation for that
domain.

 Without ontologies, or the conceptualizations that underlie knowledge, there

cannot be a vocabulary for representing knowledge.

 Thus, the first step in devising an effective knowledge-representation system,

and vocabulary, is to perform an effective ontological analysis of the field, or
domain.

 Weak analyses lead to incoherent knowledge bases.

08/08/2021 19
 Ontological Analysis (2)

 A knowledge representation language, and consequently, a

knowledge base, is built based on the ontological analysis

By
 associating terms with the concepts and relations in the ontology;

 devising a syntax for encoding knowledge in terms of the concepts and

relations

08/08/2021 20
 Ontology vs. OO (1) compare and contrast

 Building an ontology appears similar to building an Object

Oriented software

 Objects, their attributes, and their operations mirror aspects of the

domain that are relevant to the application

 OO systems representing a useful analysis of a domain can often be

reused for a different application program

08/08/2021 21
 Ontology vs. OO (2)

So what are the differences?

 Classes and objects in ontologies are about and must reflect the structure of
the world; OO objects may not.

 The representational repertoire of objects, relations, states, events, and

processes does not say anything about which classes of these entities exist.

 The modeler of the domains makes these commitments.

 Building blocks of OO are objects; building blocks of ontology are objects

and relations.

08/08/2021 22
 Ontological Analysis : Benefits

 We can share this knowledge representation language with others

who have similar needs for knowledge representation in that
domain

 Thereby eliminating the need for replicating the knowledge

analysis process

 Shared ontologies can thus form the basis for domain specific
knowledge representation languages

08/08/2021 23
 Levels of Ontology (1)

 Level 1
 Defining the concepts and the hierarchy
 e.g., Yahoo Ontology

 Level 2
 Unambiguously defining the meaning of concepts
 Adding relationship definition

 Case of many existing ontologies

 Level 3
 Task ontology

08/08/2021 24
 Levels of Ontology (2)

 As we move from an ontology top to lower taxonomic levels,

commitments specific to domains and phenomena appear

 For modeling objects on earth, we can make certain commitments

 For example, animals, minerals, and plants are subcategories of objects; has-
life(x) and contains-carbon(x) are object properties; and can-eat(x, y) is a
possible relation between any two objects

 These commitments are specific to objects and phenomena in this

domain

08/08/2021 25
 Potential Applications of ontology
Description

 Semantic integration

 Ontology driven knowledge management

 Agent mediated knowledge management

 Data and knowledge translation

 Business process modeling

 Internet information processing: Search

08/08/2021 26
 Potential Applications of ontology

 Semantic integration

 Ontology driven knowledge management

 Agent mediated knowledge management

 Data and knowledge translation

 Business process modeling

 Internet information processing: Search

08/08/2021 27
 Semantic Integration /1

 Different software systems may use individualized

conceptualizations of a certain domain.

 To achieve ontology-based semantic integration, two software

systems must find a way to share the semantics of the terms in their
ontologies, this can be done in several possible directions:
 Using a single centralized global ontology: a single centralized global ontology
is defined for the application domain and all agents or computer programs in
communication use terms from this ontology (e.g., CYC project, Experience
Factory).

08/08/2021 28
 Semantic Integration /2

Merging source ontologies into a unified ontology:

 ontologies defined on a common domain by different applications

have lots of overlap, therefore merging the source ontologies into one
unified ontology before agent interactions is a way to fulfill semantic
integration.

Mappings (or matches) between two ontologies:

 instead of trying to merge two source ontologies, finding a set of

mapping rules between them.
08/08/2021 29
 Semantic Integration /2

Runtime ontology resolution:

 Software agents may be heterogeneous and it is impractical to

restrict all agents to use a single ontology or to have ontology
merging, matching, and translation services available prior to the
deployment of the agent system.

 A better way is to resolve semantic differences when they arise

during run-time interaction

08/08/2021 30
 Agent Mediated KM

 Agent Mediated Knowledge Management (AMKM) uses agent

concepts to analyze and model organizations and their knowledge
needs and to provide a reusable architecture to build KM systems.

 AMKM is different from peer-to-peer KM in the sense that

(1) agents can learn and adapt themselves to changes in environment; and

(2) agents can get involved in complex interactions.

 These two points come from the fact that agents are autonomous
and social entities.

08/08/2021 31
 UIMA

 Unstructured Information Management (UIM) applications are

software systems that analyze large volumes of unstructured
information in order to discover knowledge that is relevant to an
end user.

 UIMA (Unstructured Information Management Architecture) is a

framework and SDK for developing such applications such as
semantic search.

08/08/2021 32
 Conclusion

 Distributed multi-agent system development relies on knowledge

bases, but those knowledge bases are built with little sharing or
reuse

 In the future, intelligent systems developers will have libraries of

ontologies. Rather than building from scratch, they will assemble
knowledge bases from components drawn from the libraries

 This should greatly decrease development time while improving the

robustness and reliability of the resulting knowledge bases
08/08/2021 33
 KQML as an Agent Communication Language
Choose

 KQML is both a message format and a message-handling protocol

to support run-time knowledge sharing among agents.

 KQML provides a basic architecture for knowledge sharing through

a special class of agent called communication facilitators which
coordinate the interactions of other agents.

08/08/2021 34
 KQML as an Agent Communication Language

 KQML is both a message format and a message-handling protocol

to support run-time knowledge sharing among agents.

 KQML provides a basic architecture for knowledge sharing through

a special class of agent called communication facilitators which
coordinate the interactions of other agents.

08/08/2021 35
 Introduction

 A community of intelligent agents can address each of the problems mentioned.

 When we describe these agents as intelligent, we refer to their ability to:

 communicate with each other using an expressive communication language;

 work together cooperatively to accomplish complex goals;

 act on their own initiative; and

 use local information and knowledge to manage local resources and handle requests from peer
agents.

 Knowledge Query and Manipulation Language (KQML) is a language that is

designed to support interactions among intelligent software agents.

 It was developed by the ARPA supported Knowledge Sharing Effort.

08/08/2021 36
Interaction: The Problem

How might meaningful, constructive and intelligent interaction

among software agents be dealt with?

 The problem for humans requires more than the knowledge of a

common language, e.g., English; it also requires a common
(mutual) understanding of the terms used in a given context.

 Example: A physicist’s understanding of velocity is not the

same as that of a F-1 driver, and if the two want to talk about
“fast cars” they have to speak a “common language.”

08/08/2021 37
1. Mutual Understanding
 Mutual understanding of what is represented may be divided into two sub
problems:
 Translation from one representation language to another (or from one family of
representation languages to another).
 Sharing the semantic content of the represented knowledge among different applications.

 Translation alone is not sufficient because each knowledge base holds implicit
assumptions about the meaning of what is represented.

 If two applications are to understand each other’s knowledge, such

assumptions must also be shared, i.e., the semantic content of the various
tokens must be preserved.

08/08/2021 38
2. Communication

Communication may be divided into three sub-problems:

 Interaction protocol: refers to the high level strategy pursued by the
software agent that governs its interaction with other agents. Such a
protocol can range from negotiation schemes and game theory
protocols, etc.
 Example: “every time you do not know something, find someone who
knows and ask.”

08/08/2021 39
2. Communication

Communication may be divided into three sub-problems:

 Communication language is the medium through which the attitudes
regarding the content of the exchange are communicated. It is the
communication language that suggests whether the content of the
communication is an assertion, a request or some form of query.
 Transport protocol is the actual transport mechanism used for the
communication, such as TCP, SMTP, HTTP, etc.

08/08/2021 40
2. Communication

To address many of the difficulties of communication among

intelligent agents, we must give them a common language.

 In linguistic terms, this means that they must share a common

syntax, semantics and pragmatics.

 Syntax: Currently, there is no universally accepted language in

which to represent information and queries.

08/08/2021 41
2. Communication

 Semantics: Assuming the use of a common language, it is still

necessary for communicating agents to share a framework of
knowledge in order to interpret the messages they exchange,
i.e., shared semantics or shared ontology.

 Pragmatics: Pragmatics among computer processes includes

1) knowing with whom to talk and how to find them; and

2) knowing how to initiate and maintain an exchange.

08/08/2021 42
How to Interact agents?

What is needed to interact and interoperate?

 A representation language KIF

 A communication language KQML

Sharing content requires more than a representation language

and a communication language

 Sharing the content Ontology

08/08/2021 43
Representation Language: KIF define
 Knowledge Interchange Format (KIF) is a language for the syntactic aspects of
representations for knowledge sharing.

 KIF can be used to support translation from one content language to another, or as
a common content language between two agents which use different native
representation languages.

 There were two intentions behind the development of KIF:

 Creation of a lingua franca for the development of intelligent applications, with an emphasis on
interoperation
 Creation of a common interchange format so that with the use of “translators” one could
translate from language A to KIF and from KIF to language B instead of translating from A to
B.

08/08/2021 44
Communication Language: KQML

 KQML was conceived as both a message format and a message-handling protocol to

support run-time knowledge sharing among agents.

 The basic features of KQML are:

 KQML messages are opaque to the content they carry. KQML messages do not merely communicate
sentences in some language, but they rather communicate an attitude about the content (assertion,
request, query).
 The language’s primitives are called performatives. Performatives define the permissible actions
(operations) that agents may attempt in communicating with each other.
 An environment of KQML speaking agents may be enriched with special agents, called facilitators,
that provide such functions as: association of physical addresses with symbolic names; registration
of databases and/or services offered and sought by agents; and communication services (forwarding,
brokering etc.). To use a metaphor, facilitators act as efficient secretaries for the agents in their
domain.
08/08/2021 45
Sharing Contents: Ontology

 Sharing content requires more than a representation language (i.e., KIF) and a
communication language (i.e., KQML).

 Every knowledge-based system relies on some conceptualization of the world

(objects, qualities, distinctions and relationships that matter for performing some
task) that is embodied in concepts, etc., in a formal representation scheme.

 Common ontologies address the problem of what must be held in common

among communicating agents.

 A common ontology refers to an explicit specification of the ontological

commitments of a set of programs. An agent commits to an ontology if its
observable actions are consistent with the definitions in the ontology.
08/08/2021 46
Knowledge Interchange Format (KIF) Define

 KIF is a version of first order predicate calculus with extensions to

support non-monotonic reasoning and definitions.

 The language description includes both a specification for its syntax and
its semantics.

 Example: expression of simple data in KIF The sentences shown below

encode 3-tuples in a personnel database (arguments stand for employee
ID number, department assignment and salary, respectively):
 (salary 015-46-3946 ENEL 72,000)

 (salary 026-40-9152 CS 36,000)

08/08/2021 47
Knowledge Interchange Format (KIF)

 KIF includes logical operators to encode logical information,

such as negation, disjunction, rules, quantified formulas and
other operators for encoding knowledge.

 KIF can also be used to describe procedures, i.e., to programs or

scripts for agents to follow. Given the syntax of KIF, such
programs resemble Lisp.

 The semantics of the KIF core (KIF without rules and

definitions) is similar to that of first order logic.
08/08/2021 48
Knowledge Query Manipulation Language (KQML)
Describe

 KQML is a language and a set of protocols that support

computer programs in identifying, connecting with and
exchanging information with other programs.

 KQML is divided into three layers:

 Content layer

 Communication layer

 Message layer

08/08/2021 49
Knowledge Query Manipulation Language (KQML)

KQML is divided into three layers:

 Content layer:
 it bears the actual content of the message, in the programs own
representation language.
 KQML can carry any representation language, including languages
expressed as ASCII strings and those expressed using a binary notation.
 Every KQML implementation ignores the content portion of the
message, except to determine where it ends

08/08/2021 50
Knowledge Query Manipulation Language (KQML)

KQML is divided into three layers:

 The communication layer:

 encodes a KQML set of features to the message

which describe the lower level communication
parameters, such as
 the identity of the sender and recipient, and

 a unique identifier associated with the communication

08/08/2021 51
Knowledge Query Manipulation Language (KQML)

KQML is divided into three layers:

 Message layer:
 is used to encode a message that one application would like to transmit to
another.
 It determines the kinds of interactions one can have with a KQML-speaking
agent.
 The prime function of the message layer is to identify the protocol to be
used to deliver the message and to supply a performative which the sender
attaches to the content (such as that it is an assertion, a query, a command,
or any of a set of known performatives).
08/08/2021 52
Knowledge Query Manipulation Language (KQML)

 Message layer:
 Since the content is opaque to KQML, this layer also includes optional
features which describe the content language, the ontology it assumes,
and some type of description of the content, such as a descriptor naming
a topic within the ontology.
 These features make it possible for KQML implementations to analyze,
route and properly deliver messages even though their content is
inaccessible.

08/08/2021 53
Communication Protocols

 There are a variety of inter-process information exchange

protocols in KQML.

 In the simplest, one agent acts as a client and sends a query to

another agent acting as a server and then waits for a reply, as is
shown between agents A and B.

 The server’s reply might consist of a single answer or a

collection or set of answers.

08/08/2021 54
Communication Protocols

 Another case, shown between agents A and C, the server’s reply

is not the complete answer but a handle which allows agent A to
ask for the components of the reply, one at a time.

 A common example of this exchange occurs when a client

queries a relational database which produces a sequence of
instantiations in response.

08/08/2021 55
Communication Protocols

 Another case, shown between agents A and D, client subscribes

to a server’s output and an indefinite number of asynchronous
replies arrive at irregular intervals.

 The client does not know when each reply message will be
arriving and may be busy performing some other task when they
do.

08/08/2021 56
Facilitators Define

 A facilitator is an agent that performs various useful

communication services

 Example: maintaining a registry of service names, forwarding

messages to named services, routing messages based on content,
providing “match-making” between information providers and
clients, and providing mediation and translation services.

 A main functions of facilitator agents is to help other agents find

appropriate clients and servers.
08/08/2021 57
Example: Facilitators

 Consider a case where an agent A would like to know the truth

of a sentence X, and agent B may have X in its knowledge-base,
and a facilitator agent F is available.

 If A is aware that it is appropriate to send a query about X to B,

then it can use a simple point to point protocol and send the
query directly to B

08/08/2021 58
Example: Facilitators

 What if A is not aware of what agents are available, or who may

have X in is knowledge-bases, or how to contact those of whom
it is aware?

 A possible scenario is that A uses the subscribe performative to

request that facilitator F monitor for the truth of X. If B
subsequently informs F that it believes X to be true, then F can
in turn inform A.

08/08/2021 59
Example: Facilitators

 A asks F to find an agent that can process an ask(X)

performative. B independently informs F that it is willing to
accept performatives matching ask(X).

 Once F has both of these messages, it sends B the query, gets a

response and forwards it to A.

08/08/2021 60
Conclusions

 The agent communication and representation languages (KQML

and KIF) and common ontologies approach together can address
the fundamental interaction and interoperation problems.

 They together include:

 Translation between representations.

 Sharing the semantic (and often pragmatic) content of the knowledge

that is represented.
 Communicating attitudes about the shared knowledge.

08/08/2021 61