Challenge@RuleML2015 Exchanging Data and Ontological Definitions in Multi-Agent-Contexts Systems
0 likes256 views
The document discusses an extension of DACMAS, a commitment-based multi-agent system, to incorporate external contexts in data-aware systems for improved data management and interoperability. It introduces a new framework, DACMACS, allowing agents to query contexts while facilitating non-monotonic queries through bridge rules for tailored data extraction. The proposed system enhances knowledge exchange by allowing agents to share ontological definitions and manage updates for global conformity while ensuring runtime verifiability.
![Bridge Rules in a DACMACS
A(ˆx) determinedby E1, . . . , Ek , not Gk+1, . . . , not Gr
where each of the Eis and the Gis are datalog
queries to external context, whose reference is either
locally known or provided by the Institutional agent
via a query Role@inst(role)
A bridge rule is proactively enabled by a trigger rule:
Q(ˆx) enables A(ˆy) [Time | Frequency]
Then, results are exploited via a bridge-update rule:
upon A(ˆx) then β(ˆx)
where β(ˆx) specifies the operator, management
function and actions to be applied to ˆx.](https://image.slidesharecdn.com/challengeruleml2015exchangingdataandontologicaldefinitionsinmulti-agent-contextssystems-150806082350-lva1-app6892/85/Challenge-RuleML2015-Exchanging-Data-and-Ontological-Definitions-in-Multi-Agent-Contexts-Systems-10-320.jpg)
Challenge@RuleML2015 Exchanging Data and Ontological Definitions in Multi-Agent-Contexts Systems
- 1. Exchanging Data and Ontological Definitions in Multi-Agent-Contexts Systems Stefania Costantini Giovanni De Gasperis Dipartimento di Ingegneria e Scienze dell’Informazione e Matematica University of L’Aquila, Italy RuleML 2015, Challenges Track 5 August 2012, Berlin, DE
- 2. Exchanging Data and Ontological Definitions in Multi-Agent-Contexts Systems Abstract We have extended DACMAS, a formalization of ontology-based data-aware and commitment-based multi-agent systems. The extension allows a system to include not only agents but also external contexts. The aim is to model real-world situations where agents not only interact among themselves, but also consult external heterogeneous data- and knowledge-bases to extract useful information. In this work we further enhance the approach, so that a querying agent is enabled to specify in ontological terms which data it intends to extract from a context, or vice versa the received results require the agent to be aware of ontological assumptions they are based upon.
- 3. Data Representation and Knowledge Exchange in Agents The importance of data/knowledge representation and exchange in Artificial Intelligence applications is constantly increasing. In many application fields it is particularly important to comprise and elaborate information provided by multiple sources. Logic-based data management and exchange are therefore important issues in logical agents. Non-monotonic reasoning both in defining and executing patterns for knowledge exchange and in the modalities for knowledge exploitation.
- 4. Tboxes and Aboxes (Courtesy of http://ai.ia.agh.edu.pl)
- 5. DACMAS: Data-Aware Commitment-based Multi-Agent System (MAS) Introduced by Montali et al, 2014 A recent interesting data-centric agent architecture. Institutional agent which owns a “global” (DLR-Lite) 1 TBox, specifying the domain in which the MAS operates. Each participating agent is equipped with its local ABox (consistent with the TBox, mutual consistency not required). Commitment-based Communication. (example: buyer agent generating orders became customer, then seller delivers) Patterns for simple reactive and proactive rules. 1 n-ary version of the DL at the base of the OWL 2 QL profile
- 6. Our view of DACMAS DACMAS: yet another agent architecture or useful Meta-Model for Multi-Agent Systems? Particularly interesting for a general specification of data management and communicative features. Allows for affordable verification (cit. Montali et al). Very general about an agent program’s definition, so it can be specialized to many existing agent-oriented logic languages, including our logic framework DALI. Missing: access to external knowledge sources. (this is one of our proposed extension)
- 7. Accessing External Sources from MAS Options available: Agents & Artifacts approach: postulates that a (homogeneous) description is available for such sources, that can in general be manipulated by agents. (managed) Multi-Context Systems (mMCSs): drop the assumption of making external sources in some sense homogeneous: rather, the approach deals explicitly with their different representation languages and semantics. mMCSs: allow for datalog-like non-monotonic queries to external sources, called “bridge rules”.
- 8. Our first extension: DACMACS Data-Aware Commitment-based managed Multi-Agent-Context Systems Integrates DACMASs and mMCS: Agents can query (sets of) contexts, but contexts cannot query agents. Agents are equipped with bridge rules, whose application is however activated via special trigger rules, which allow a bridge rule to be invoked upon certain conditions and/or according to a certain timing. The result of a bridge rule is interpreted as an agent-generated internal event, and captured by reactive rules which may determine modifications to the agent’s ABox.
- 9. Bridge Rules in mMCSs how they look like in general form o(s) ← (c1 : p1), . . . , (cj : pj), not (cj+1 : pj+1), . . . , not (cm : pm). where the cis are contexts, i.e., external knowledge sources, and o is a knowledge integration operator dealt with by a management function
- 10. Bridge Rules in a DACMACS A(ˆx) determinedby E1, . . . , Ek , not Gk+1, . . . , not Gr where each of the Eis and the Gis are datalog queries to external context, whose reference is either locally known or provided by the Institutional agent via a query Role@inst(role) A bridge rule is proactively enabled by a trigger rule: Q(ˆx) enables A(ˆy) [Time | Frequency] Then, results are exploited via a bridge-update rule: upon A(ˆx) then β(ˆx) where β(ˆx) specifies the operator, management function and actions to be applied to ˆx.
- 11. Semantics of a DACMACS Mapped to DALI implementation DACMAS: operational and asynchronous transition system, native infrastucture in DALI. 2 mMCSs equilibria : global acceptable data states, one for each context, guided by bridge rules for inter-context communication, actuated by means of operational statements in management functions. Bridge rules are deemed to be applied whenever applicable. 3 DACMACS equilibria : they are extended mMCSs because of the introduction of agents and the bridge rule applicability implied by agents’ proactive choice. 2 DALI framework supports asynchronous external and internal events with logic programming defined behaviors of agents 3 DALI precondition rules associated to internal events
- 12. Properties of a DACMACS MAS verification Safe evolution trajectory of a DACMACS: sequence of equilibria w.r.t. agents’ and contexts’ knowledge base updates. Local Consistency: equilibria are composed of consistent data sets. A DACMACS (or an mMCS) enjoys Local Consistency iff all management functions are local consistency (lc-) preserving. Global Consistency not required.
- 13. DACMACS benefits DACMACS = DACMAS + mMCS = Data-centric agents + knowledge exchange with external knowledge bases Issue: a-priori and run-time verification Extension: Agents and Contexts exchange Ontological Definitions Extension: use acquired ontological definitions for positive and negative explanations, remember the sources of such definitions, update their level of trust and their reputation.
- 14. Yet an other extension: tractable ontological query-answering How to interoperate at ontological level Agents have their own TBox and can generate ontological queries to the institutional agent a global ontology is available to agents and contexts to achieve interoperability (example: a shared taxonomy, a dictionary..) the union of global and local TBoxes is consistent the local TBox can be updated via bridge rules the global TBox in a DACMACS is a protected fragment from agents local TBox updates new definitions derived from an ontological answer can be added to the local TBox preserving consistency
- 15. Example: Student enrollment Check for eligibility to enroll In a DACMACS representing the University, the agent "student_secretariat" may consult the contexts "student_office" for grades and "tax_office" for family income. bridge rule: eligible(stud) determinedby low_income(stud) : tax_office, not nonexcellent(stud) : student_office low_income(stud) :- student(stud), (1) family_income(stud, inc), inc ≤ inc_threshold. nonexcellent(stud) :- student(stud), (2) grade(stud, grad), grad ≤ grade_threshold.
- 16. Example: Student enrollment more general ontology based query eligible(stud) determinedby low_income(stud) : tax_office : D1, not nonexcellent(stud) : student_office : D2 where D1 and D2 are variables initiated by respective office contexts to (1) and (2)
- 17. Extended bridge Rules A(ˆx) determinedby E1, . . . , Ek , not Gk+1, . . . , not Gr where each of the Eis and the Gis are datalog queries to external context, whose reference is either locally known or provided by the Institutional agent via a query with variables parameters that can be instantiated at context level We assume that all variables occurring in A(ˆx) and in each of the Gis also occur in the Eis. Within an agent, different bridge rules have distinct conclusions. The management operations and function are defined separately. Satisfiability Preservation: Updates should preserve satisfiability of basic concepts and roles. Protection: Updates should preserve the protected fragment of the TBox.
- 18. Conclusions Extended DACMACS allows a system interoperability between agents and external contexts external contexts can return not only grounded query answers, but also their ontological definitions the overall system is able to evolve by incorporating new data/knowledge, and new ontological information, maintaining verifiability at run-time. the model is currently being mapped to a DALI/Python based implementation with logical agents and procedural institutional agent
- 19. The End! Thank You for your Attention:-) Questions? stefania.costantini@univaq.it giovanni.degasperis@univaq.it