A Novel Service Oriented Model for Query Identification and Solution Development using Semantic Web and Multi Agent System

In this paper, we propose to develop service model architecture by merging multi-agentsystems and semantic web technology. The proposed architecture works in two stages namely, Query Identification and Solution Development. A person referred to as customer will submit the problem details or requirements which will be referred to as a query. Anyone who can provide a service will need to register with the registrar module of the architecture. Services can be anything ranging from expert consultancy in the field of agriculture to academic research, from selling products to manufacturing goods, from medical help to legal issues or even providing logistics. Query submitted by customer is first parsed and then iteratively understood with the help of domain experts and the customer to get a precise set of properties. Query thus identified will be solved again with the help of intelligent agent systems which will search the semantic web for all those who can find or provide a solution. A workable solution workflow is created and then depending on the requirements, using the techniques of negotiation or auctioning, solution is implemented to complete the service for customer. This part is termed as solution development. In this service oriented architecture, we first try to analyze the complex set of user requirements then try to provide best possible solution in an optimized way by combining better information searches through semantic web and better workflow provisioning using multi agent systems.

💡 Research Summary

The paper proposes a hybrid service‑oriented architecture that integrates Semantic Web technologies with a Multi‑Agent System (MAS) to address the problem of translating complex user requirements into concrete service solutions. The workflow is divided into two principal phases: Query Identification and Solution Development.

In the Query Identification phase, a “customer” submits a textual description of a problem, which is first parsed by a natural‑language module and then mapped onto a domain ontology expressed in OWL‑DL. Because initial mappings are often incomplete or ambiguous, the system engages domain‑expert agents in an iterative question‑answer loop with the customer. This interaction refines the query into a precise set of attribute‑value pairs, effectively converting an unstructured request into a structured semantic representation.

The Solution Development phase begins once the query is fully specified. Service providers must register with a central registrar, publishing their service descriptions as RDF triples on the Semantic Web. A search agent issues SPARQL queries against this knowledge base, retrieving candidate providers that match both functional and non‑functional criteria (e.g., expertise, QoS, price). The candidates are then filtered using multi‑criteria decision‑making techniques.

When multiple suitable providers exist, the architecture invokes negotiation and auction agents. Negotiation follows the FIPA‑ACL protocol to establish a Service Level Agreement (SLA) acceptable to both parties. Simultaneously, an auction agent can run a reverse or multi‑round auction to select the provider offering the best cost‑benefit ratio. The outcome of this market‑driven process is a contract that is automatically instantiated.

A workflow generation module then constructs a BPMN‑like process model that orchestrates the selected provider’s tasks, data flows, and verification steps. This model is executed by a workflow engine, delivering the final service to the customer with minimal human intervention.

The authors claim three main contributions: (1) a semantic‑driven query refinement mechanism that improves requirement precision compared with traditional keyword‑based matching; (2) a modular MAS that delegates searching, matching, negotiation, and auctioning to specialized agents, enhancing scalability and maintainability; and (3) the integration of market‑based auction mechanisms to achieve optimal resource allocation and cost reduction.

Nevertheless, the paper leaves several open issues. The cost and effort required to build and maintain comprehensive ontologies are not quantified, and the experimental evaluation lacks large‑scale performance metrics such as latency and throughput under heavy load. Trust and security between agents—authentication, authorization, and data integrity—are only briefly mentioned, raising concerns for real‑world deployment. Moreover, potential strategic behaviors in auctions (e.g., collusion) are not addressed, suggesting the need for robust auction design.

In summary, the work presents an ambitious framework that combines the expressive power of the Semantic Web with the autonomy and negotiation capabilities of MAS to transform vague user queries into optimized service workflows. Future research directions include automated ontology generation, robust trust frameworks for agents, and scalable, strategy‑proof auction algorithms to strengthen the practicality of the proposed system.