The LARKS Project

1
The LARKS Project
Interoperability among Heterogeneous Agents in
the Internet
Katia Sycara, Matthias Klusch, Jianguo Lu,
Zhendong Niu, Seth Widoff
http//www.cs.cmu.edu/softagents/interop/matchmak
ing.html
2
Language for Advertisement and Requests
for Knowledge Sharing
What is LARKS about?

• Frame-based specification of ads and requests
  of agent capabilities.
• Optional use of a domain ontology.
• Automated processing of LARKS specifications.
• Compatibility with XML.

• Matchmaking Services via Matchmaker Agents
  using LARKS

• to match requests with relevant provider agents

in heterogeneous and dynamic agent societies.
3
Matchmaking Using LARKS
Matchmaker Agent x
AdvertisementDB OntologyDB AuxiliaryDB
Matching
Result-of-Matching
Capability Descriptions in LARKS
Requester Agent
Provider Agent 1
Service Request in LARKS
LARKS Protocol for providing the
service
Provider Agent n
Local Ontology 1
IS
IS
?
Process Request on Local IS
IS
Local Ontology n
4
Specification in LARKS

• Context of Specification

Context
Type

• Data Types Used in Variable Declarations

Input

• Input/Output Variable Declarations

Output
InConstraints OutConstraints

• Logical Constraints on Variables

(Pre-/Post-Conditions)
ConcDescriptions

• Ontological Description of Used Words

TextDescription

• Textual Description of Specification

• Constraints set of definite program
  clauses/goals (Horn clauses)
• Concepts ontological description of words in
  concept language ITL

5
Example for Ontology
(1) An Ontology written in the Concept Language
ITL (Terminology)
AirMission ? (and Mission (atleast 1
had-airplane) (all has-airplane Airplane)
(all has-MissionType
aset(AWAC,BARCAP,CAP,DCA,HVAA)) )
AWAC-AirMission ? (and AirMission (atleast 1
has-airplane) (atmost 1 has-airplane)
(all has-airplane
aset(E-2)) )
DCA-AirMission ? (and AirMission (all
has-MissionType aset(DCA))
(all has-F14 plane-F14D) (all
has-F18 plane-F18)
(atleast 2 has-F14) (atmost 2 has-F14)
(atleast 2 has-F18)
(atmost 2 has-F18) (atleast 2 has-E2)
(atmost 2 has-E2) )
. . .
(2) Concept Subsumption Hierarchy
AirMission
. . .
AWAC-AirMission DCA-AirMission
HVAA-AirMission
6
Examples for Specification in LARKS
Agent Advertisement I can provide information
about air combat missions in general.
AirCombatMissions
Context
Attack, Air, Combat, MissionAirMission
Types
Date (mm Int, dd Int, yy Int)
Input
Output
missionTypes SetOf(String) missionAirplanes
SetOf(String) missions ListOf(mType String,
mIDStringInt, mStart
Date, mEnd Date)
InConstraints
mStart lt mEnd
OutConstraints
ConcDescriptions
AirMission
information about air combat missions in general
TextDescription
7
Agent Advertisement I can provide information
about deployed and launched AWAC air combat
missions.
AWAC-CombatMissions
Context
Combat, MissionAWAC-AirMission
Types
Date (mm Int, dd Int, yy Int)
DeployedMission ListOf(mType String,
mIDStringInt, mStart Date, mEnd
Date)
Input
Output
missions DeployedMission
InConstraints
OutConstraints
deployed(mID), mType AWAC, launched_after(mID,m
Start), launched_before(mID,mEnd).
ConcDescriptions
AWAC-AirMission
TextDescription
information about deployed and launched AWAC air
combat missions
8
Agent Advertisement I can provide information
about deployed and launched DCA air combat
missions.
DCA-CombatMissions
Context
Combat, MissionDCA-AirMission
Types
Date (mm Int, dd Int, yy Int)
DeployedMission ListOf(mType String,
mIDStringInt, mStart Date, mEnd
Date)
Input
Output
missions DeployedMission
InConstraints
deployed(mID), mType DCA, launched_after(mID,mS
tart), launched_before(mID,mEnd).
OutConstraints
ConcDescriptions
DCA-AirMission
TextDescription
information about deployed and launched DCA air
combat missions
9
Request Find an agent who can provide
information on deployed air combat missions
launched in a given time interval.
Req-AirCombatMissions
Context
Attack, MissionAirMission
Types
Date (mm Int, dd Int, yy Int) Mission
ListOf(mType String, mIDStringInt,
mStart Date, mEnd Date)
Input
sd Date, ed Date
Output
missions Mission
InConstraints
sd lt ed.
OutConstraints
deployed(mID), launched_after(mID,sd),
launched_before(mID,ed), sd gt mStart, ed lt
mEnd.
ConcDescriptions
AirMission
information on deployed air combat missions
launched in a given time interval
TextDescription
10
Request Find an agent who can provide
information on deployed and launched air
combat missions that contain F-14 airplanes.
Req-DeployedF14CombatMissions
Context
Combat, MissionF14-Mission
Types
Date (mm Int, dd Int, yy Int) Mission
ListOf(mType String, mIDStringInt,
mStart Date, mEnd Date)
Input
Output
missions Mission
InConstraints
OutConstraints
deployed(mID), launched_after(mID,mStart), launche
d_before(mID,mEnd).
ConcDescriptions
F14-Mission ? (and AirMission (atleast 1 has-F14))
information on deployed and launched air combat
missions that contain F-14 airplanes
TextDescription
11
An Example of Matchmaking Using LARKS
Find agents that can inform about air combat
missions
Requester Agent
Ranked Set of Agents capable to
inform about air combat missions
AirCombatMissions
Matchmaker Agent
AuxiliaryDB
OntologyDB
AdvertisementDB
AWAC-AirCombatMissions
Context Matching
Signature Matching
Semantical Matching
Similarity Matching
Profile Matching
12
Matchmaking Process
Main Steps
(pairwise matching of specifications in LARKS)

• Context Matching

• Word Distances
• Subsumption or Equality of attached Concepts

• Syntactical Matching

• Comparison of Profiles (TF-IDF)

• Similarity Matching of Declarations and
  Constraints

• Word Distances
• Distances in the Ontology (Weighted Associative
  Network)

• Full Signature Matching of Declarations

• Subsumption of Data Types

• Semantical Matching

• Logical Implication of Constraints

13
Different Matching Modes
? Complete Matching All
filters. ? Relaxed Matching No
signature matching. ? Profile Matching
Comparison of profiles (TF-IDF) only. ?
Plug-In Matching Signature and
semantical matching only.
Any combination of filters may be chosen by the
user.
14
Matchmaking Process Simple Example
? Request for Agent Capability in
Req-AirCombatMissions Advertisement of
Agent Capability in AWAC-AirCombatMissions
Mode ? Complete Matching

• Context Matching ?

Look-up/compute word distances for (Attack,
Combat), (mission, Mission) All value exceeds
given threshold. Attached concepts
(AirMission,AirMission) are terminologically
equal.

• Syntactical Matching

• Comparison of Profiles (TF-IDF) ?

Profile similarity value of documents exceeds
given threshold. Thus, proceed with
15
Matchmaking Process Simple Example (contd)

• Syntactical Matching (ff.)

• Full Signature Matching of Declarations ?

Input declarations - A D21 sd Date, ed Date
Output declarations - R D12 missions
ListOf(mType String, mIDStringInt,
mStart Date, mEnd Date) A
D22 missions ListOf(mType String,
mIDStringInt,
mStart Date, mEnd Date) A D23 missionTypes
ListOf(String) A D24 missionAirplanes
ListOf(String)

• Apply Type Inference Rules for Computing Type
  Subsumption.

Result (TRUE) (signature(A D22)
signature(R D12)) Thus, proceed with
16
Matchmaking Process Simple Example (contd)

• Similarity Matching ?

• Conceptual Attachment to words

• Check word distances in declarations and
  constraints

Similarity value of two specifications Average
of sum of similarity computations among all
pairs of declarations and constraints. Similarity
of AirCombatMissions and
Req-AirCombatMissions
0.75 (Note Value strongly depends on the
used word distance database)
Similarity value exceeds given threshold. Thus,
proceed with
17
Matchmaking Process Simple Example (contd)

• Constraint Matching
  fails

InConstraints ?
InConstraints in Req-AirCombatMissions
in AirCombatMissions ? sd lt ed.

OutConstraints ?
OutConstraints NOT ? in
AirCombatMissions in
Req-AirCombatMissions mStart lt mEnd
deployed(mID),
launched_after(mID,sd),

launched_before(mID,ed), sd gt mStart, ed lt mEnd.
18
Matchmaking Process Simple Example (2)
? Request Req-DeployedF14CombatMissions, Ad
DCA-CombatMissions

• Context Matching ? (Ad) DCA-AirMission lt
  (Req) F14-Mission

• Syntactical Matching

• Comparison of Profiles (TF-IDF) ?

• Full Signature Matching of Declarations ?

• Similarity Matching ? Sim(Req,Ad)
  0.62

• Constraint Matching

InConstraints in Req ?
InConstraints in Ad ?
OutConstraints in Ad ?
OutConstraints in Req ?
deployed(mID), mTypeDCA, launched_after(mID,sd),
launched_before(mID,ed),
deployed(mID), launched_after(mID,sd),
launched_before(mID,ed),
19
Current Implementations of LARKS Project

• Web-based Interface for Specification of
  Requests in LARKS
• Matchmaker module
• Knowledge management module (Concept
  classifier, KB)

20
Graphical User Interface to the Matchmaker
21
TSC Concept Classifier for Knowledge Base
22
Matchmaking Using LARKS and XML

• Translation of LARKS specification into XML
  document
• Automated Processing of LARKS/XML Specs

• Translation of ITL concept definitions into XML
  statements

e.g. use of XML-Data Schemas, W3C Note 05 Jan
1998 Proposal by Microsoft http//www.w3.org/TR/
1998/NOTE-XML-data-0105/

• Translated ITL ontologies in XML used as
  XML-namespaces

. . . .
23
Some Future Work of the LARKS Project

• Complete Implementation in Java
• Multi-Agent Test Environment Using RETSINA
  Communicator
• Natural Language Interface for Human-Agent
  Interaction
• Extensive Real-World Testing