Towards a satisfactory conversion of messages among agent-based information systems
Highlights
► A framework for seamless communication among heterogeneous agent systems is presented. ► Message conversion is facilitated by an ontology of communication acts. ► Proposed conversions are checked by an Event Calculus representation of the scenario. ► A prototype allows us to observe the behavior of the framework in specific scenarios.
Introduction
Nowadays, the vision of isolated information systems that work on their own without cooperating with other systems is no longer realistic. Information systems which have been independently developed by different organizations need to communicate with each other in order to enhance their functionality. However, several problems must be solved before real communication is achieved. Although there are several levels where communication issues may occur, such as the transport level or the application protocol level, in this paper we focus only on how to achieve utterance communication at a semantic level.
We adopt the cooperative software agents approach as communication platform since agent technology has been proved useful for solving problems with a highly distributed nature that need flexible and adaptable solutions (e.g. (Agogino and Tumer, 2012, Sato et al., 2012)). Communication between software agents is based on the interchange of messages. When an agent from one system wants to communicate with an agent of another system, it generates a message following the rules that have been established in the former system in order to write messages. These rules imply the use of fixed structures, languages and meaning, which is useful within a certain agent system but becomes useless in the majority of situations when trying to communicate between agents from heterogeneous systems. We advocate the use of a more flexible interoperation where the interpretation of a message is made dynamically and transparently for the heterogeneous agents that want to communicate with each other.
In the relevant literature, several works can be found that consider the problem of how to achieve a flexible communication between agents from heterogeneous systems. Details of this are presented in Section 6 but, in summary, we can say that a high number of solutions point towards the development of mechanisms that only deal with syntactical aspects of agent communication languages (ACLs) (see (Lopes and Botelho, 2005, Suguri et al., 2002)). Furthermore, a trend of works emphasize the adequacy of communication protocols in order to restrict the potentially wild set of utterance choices. Models of human dialogues have been used to categorize some types of dialogues (Walton & Krabbe, 1995) and protocol designers have adopted such source as a basis for structuring agent communications (Maudet & Chaib-Draa, 2002). Analyses have been undertaken on desiderata for such protocols (McBurney, Parsons, & Wooldridge, 2002). Doubtless, communication protocols are an important part of the play of software agent interactions, and they are subject to a lot of study and effort (Berges et al., 2011, Chopra and Singh, 2009, Desai et al., 2009). Nevertheless, heterogeneity of already existing information systems exhibits great differences in the structure and intended semantics of their messages. Therefore it is not guaranteed that they satisfy the basic assumptions considered in some protocol models (i.e. semantics of individual messages expected in the protocol do not match with the message semantics of the participating agent). In our opinion, semantic technologies such as ontologies can be used in order to handle semantic aspects properly, and in particular, we advocate for managing semantics of individual messages in order to increase the possibilities of success when dealing with protocols. Other authors share the opinion that the introduction of semantic-based technology in information systems provides for a significant enhancement of their functionalities and capabilities. For example authors in (Shaw, Tsao, & Wu, 2012) study the application of ontologies in order to create FAQ auto-categorizing systems that can improve customer service in technical support centers and alike. Moreover, in (Chen, Huang, Bau, & Chen, 2012) ontologies and the Semantic Web Rule Language (SWRL1) are used in order to develop a diabetes medication recommendation system. This system aims at suggesting a prescription for a patient based on knowledge about symptoms and diabetes-related drugs. Finally authors in (Belmonte, Pérez-de-la-Cruz, & Triguero, 2008) describe an ontology-based multiagent decision support system for bus fleet management, whose goal is to reason about traffic behavior in the same way an expert traffic operator would. However, as opposed to our approach, only agents that use the same ACL are considered. In this sense, we present a framework consisting of a communication acts ontology, called CommOnt, a translation mediator, an Event Calculus axiomatization of the scenario and some agents in charge of managing the communication process between systems.
In summary, the main features of the proposed framework are:
- •
It favors a semantic communication between heterogeneous and distributed information systems using semantic technologies and therefore it increases their cooperation opportunities.
- •
It incorporates an ontology that describes the types of communication acts used by software agents. This ontology allows the recognition of instances of communication acts from one system as instances of communication acts of another system.
- •
It also integrates domain and action ontologies with the goal of extending its applicability to a wide range of different scenarios.
- •
It incorporates Event Calculus sentences to represent the scenario. Model theory is used to formalize a notion of satisfactory conversion. Event Calculus reasoners allow to check proposed conversions.
- •
A developed prototype allows us to observe the behavior of the framework in specific scenarios. In particular, in this paper, the feasibility of our framework is presented in a scenario where one agent of our MedicalFIPAAgents system communicates with one agent of the Aingeru system, two healthcare information systems which use different languages, structures and meanings in the composition of their messages.
The rest of the paper is organized as follows: In Section 2 the main components of the proposed framework are presented, as well as a brief description of the communication process. Section 3 explains preliminary notions about the ontologies in the framework and the dynamic interpretation of communication acts. Section 4 introduces the formalization of the notion of satisfactory conversion, while one scenario of the proposed framework at work is presented in Section 5. Related works are discussed in Section 6 and finally, we end with some conclusions in Section 7.
Section snippets
Global architecture of the framework
In this section the architecture of the framework is presented from a global point of view. First, the elements that belong to the framework are briefly described, emphasizing the role that each element plays in the architecture. Then, the process applied to one message from the moment it is created until it reaches its final destination is shown.
Preliminaries
In this section preliminary notions about the ontologies in the framework and our adopted approach for the dynamic interpretation of communication acts are explained.
Formalization of satisfactory conversion
Once the conversion process has finished, it is reasonable to ask whether the communicating agents agree on the understanding of the sent and received message. It is conceivable that, although their respective message classes are related properly in the communication acts ontology, their intended effects may not match accurately (i.e. the semantics of the sent message is not exactly the same as the semantics of the received message).
In this section, first, the description of the conversion
The proposed framework at work
In this section one scenario that illustrates the different steps that are followed by the proposed framework to achieve semantic communication between software agents from two different information systems is presented. System MedicalFIPAAgents is composed of agents that use FIPA-ACL as agent communication language and FIPA-SL0 (FIPA-SL, 2002) as content language at the time of composing messages. Moreover, system Aingeru is composed of agents that use specific communication acts in their
Related works
Several works that treat the topic of interoperability among agents from heterogeneous systems can be found in the specialized literature. The work of Lopes and Botelho (2005) suggests the idea of translating the content of every message to an abstract logic language defined by the authors (and vice versa). This work is similar to ours because it makes use of an intermediate language. Nevertheless, it only takes into account the syntax of the messages, so it allows interoperability between
Conclusions
We have introduced a proposal, based on Semantic Web technology, to tackle the problem of semantic communication among heterogeneous and distributed information systems represented by software agents. A novel feature of that proposal is that it favors a flexible communication process among those systems avoiding a priori agreements about interchanged messages. Such flexibility is obtained by considering the semantics associated to the communication process. The proposed framework includes as
Acknowledgements
The work of Idoia Berges was supported by a grant of the Basque Government (Programa de Formación de Investigadores del Departamento de Educación, Universidades e Investigación). This work is also supported the Spanish Ministry of Education and Science TIN2010–21387-C02–01.
References (50)
- et al.
Ontologies and agents for a bus fleet management system
Expert Systems with Applications
(2008) - et al.
A recommendation system based on domain ontology and SWRL for anti-diabetic drugs selection
Expert Systems with Applications
(2012) Circumscription – A form of non-monotonic reasoning
Artificial Intelligence
(1980)- et al.
A study of the application of ontology to an FAQ automatic classification system
Expert Systems with Applications
(2012) - et al.
A multiagent approach to managing air traffic flow
Autonomous Agents and Multi-Agent Systems
(2012) - Berges, I., (2011). Semantic interoperability among heterogeneous information systems: Two use cases. PhD Thesis,...
- et al.
A mechanism for discovering semantic relationships among agent communication protocols
Autonomous Agents and Multi-Agent Systems
(2011) - Boella, B., Damiano, R., Hulstijn, J., & van der Torre, L. W. N. (2006). Role-based semantics for agent communication:...
- Boella, B., Hulstijn, J., & van der Torre, L. W. N. (2009). Social network semantics for agent communication. In 8th...
Amoeba: A methodology for modeling and evolving cross-organizational business processes
ACM Transactions on Software Engineering Methodology
Defining interaction protocols using a commitment-based agent communication language
Specifying artificial institutions in the event calculus
Representation and monitoring of commitments and norms using OWL
AI Communications
Agent communication and artificial institutions
Autonomous Agents and Multi-Agent Systems
Agent interoperation across multiagent system boundaries
Cited by (5)
Modeling and verifying probabilistic Multi-Agent Systems using knowledge and social commitments
2014, Expert Systems with ApplicationsA user's cognitive workload perspective in negotiation support systems: An eye-tracking experiment
2016, Pacific Asia Conference on Information Systems, PACIS 2016 - ProceedingsBinding SNOMED CT terms to archetype elements: Establishing a baseline of results
2015, Methods of Information in MedicineUser acceptance in different electronic negotiation systems - A comparative approach
2013, Proceedings - 2013 IEEE 10th International Conference on e-Business Engineering, ICEBE 2013